CN112861907A - Method for tracing origin of white tea - Google Patents

Method for tracing origin of white tea Download PDF

Info

Publication number
CN112861907A
CN112861907A CN202011643531.4A CN202011643531A CN112861907A CN 112861907 A CN112861907 A CN 112861907A CN 202011643531 A CN202011643531 A CN 202011643531A CN 112861907 A CN112861907 A CN 112861907A
Authority
CN
China
Prior art keywords
white tea
sample
samples
calculating
producing areas
Prior art date
Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
Pending
Application number
CN202011643531.4A
Other languages
Chinese (zh)
Inventor
孙威江
林刚
罗玉琴
周嘉羿
商虎
林孝潮
Current Assignee (The listed assignees may be inaccurate. Google has not performed a legal analysis and makes no representation or warranty as to the accuracy of the list.)
Fujian Rongyuntong Ecological Technology Co ltd
Fujian Agriculture and Forestry University
Original Assignee
Fujian Rongyuntong Ecological Technology Co ltd
Fujian Agriculture and Forestry University
Priority date (The priority date is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the date listed.)
Filing date
Publication date
Application filed by Fujian Rongyuntong Ecological Technology Co ltd, Fujian Agriculture and Forestry University filed Critical Fujian Rongyuntong Ecological Technology Co ltd
Priority to CN202011643531.4A priority Critical patent/CN112861907A/en
Publication of CN112861907A publication Critical patent/CN112861907A/en
Pending legal-status Critical Current

Links

Images

Classifications

    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06F—ELECTRIC DIGITAL DATA PROCESSING
    • G06F18/00—Pattern recognition
    • G06F18/20—Analysing
    • G06F18/21—Design or setup of recognition systems or techniques; Extraction of features in feature space; Blind source separation
    • G06F18/213—Feature extraction, e.g. by transforming the feature space; Summarisation; Mappings, e.g. subspace methods
    • G06F18/2132—Feature extraction, e.g. by transforming the feature space; Summarisation; Mappings, e.g. subspace methods based on discrimination criteria, e.g. discriminant analysis
    • G—PHYSICS
    • G01—MEASURING; TESTING
    • G01N—INVESTIGATING OR ANALYSING MATERIALS BY DETERMINING THEIR CHEMICAL OR PHYSICAL PROPERTIES
    • G01N21/00—Investigating or analysing materials by the use of optical means, i.e. using sub-millimetre waves, infrared, visible or ultraviolet light
    • G01N21/17—Systems in which incident light is modified in accordance with the properties of the material investigated
    • G01N21/25—Colour; Spectral properties, i.e. comparison of effect of material on the light at two or more different wavelengths or wavelength bands
    • G01N21/31—Investigating relative effect of material at wavelengths characteristic of specific elements or molecules, e.g. atomic absorption spectrometry
    • G01N21/35—Investigating relative effect of material at wavelengths characteristic of specific elements or molecules, e.g. atomic absorption spectrometry using infrared light
    • G01N21/3563—Investigating relative effect of material at wavelengths characteristic of specific elements or molecules, e.g. atomic absorption spectrometry using infrared light for analysing solids; Preparation of samples therefor
    • G—PHYSICS
    • G01—MEASURING; TESTING
    • G01N—INVESTIGATING OR ANALYSING MATERIALS BY DETERMINING THEIR CHEMICAL OR PHYSICAL PROPERTIES
    • G01N21/00—Investigating or analysing materials by the use of optical means, i.e. using sub-millimetre waves, infrared, visible or ultraviolet light
    • G01N21/17—Systems in which incident light is modified in accordance with the properties of the material investigated
    • G01N21/25—Colour; Spectral properties, i.e. comparison of effect of material on the light at two or more different wavelengths or wavelength bands
    • G01N21/31—Investigating relative effect of material at wavelengths characteristic of specific elements or molecules, e.g. atomic absorption spectrometry
    • G01N21/35—Investigating relative effect of material at wavelengths characteristic of specific elements or molecules, e.g. atomic absorption spectrometry using infrared light
    • G01N21/359—Investigating relative effect of material at wavelengths characteristic of specific elements or molecules, e.g. atomic absorption spectrometry using infrared light using near infrared light
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06F—ELECTRIC DIGITAL DATA PROCESSING
    • G06F18/00—Pattern recognition
    • G06F18/20—Analysing
    • G06F18/24—Classification techniques
    • G06F18/241—Classification techniques relating to the classification model, e.g. parametric or non-parametric approaches
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06F—ELECTRIC DIGITAL DATA PROCESSING
    • G06F18/00—Pattern recognition
    • G06F18/20—Analysing
    • G06F18/24—Classification techniques
    • G06F18/241—Classification techniques relating to the classification model, e.g. parametric or non-parametric approaches
    • G06F18/2415—Classification techniques relating to the classification model, e.g. parametric or non-parametric approaches based on parametric or probabilistic models, e.g. based on likelihood ratio or false acceptance rate versus a false rejection rate
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
    • G06N3/00—Computing arrangements based on biological models
    • G06N3/02—Neural networks
    • G06N3/04—Architecture, e.g. interconnection topology
    • G06N3/047—Probabilistic or stochastic networks
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
    • G06N3/00—Computing arrangements based on biological models
    • G06N3/02—Neural networks
    • G06N3/04—Architecture, e.g. interconnection topology
    • G06N3/048—Activation functions
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
    • G06N3/00—Computing arrangements based on biological models
    • G06N3/02—Neural networks
    • G06N3/08—Learning methods
    • G06N3/084—Backpropagation, e.g. using gradient descent

Landscapes

  • Engineering & Computer Science (AREA)
  • Physics & Mathematics (AREA)
  • Theoretical Computer Science (AREA)
  • Data Mining & Analysis (AREA)
  • General Physics & Mathematics (AREA)
  • Life Sciences & Earth Sciences (AREA)
  • Evolutionary Computation (AREA)
  • General Engineering & Computer Science (AREA)
  • Artificial Intelligence (AREA)
  • General Health & Medical Sciences (AREA)
  • Health & Medical Sciences (AREA)
  • Computer Vision & Pattern Recognition (AREA)
  • Spectroscopy & Molecular Physics (AREA)
  • Computational Linguistics (AREA)
  • Biomedical Technology (AREA)
  • Mathematical Physics (AREA)
  • Software Systems (AREA)
  • Molecular Biology (AREA)
  • Bioinformatics & Cheminformatics (AREA)
  • Bioinformatics & Computational Biology (AREA)
  • Biophysics (AREA)
  • Evolutionary Biology (AREA)
  • Computing Systems (AREA)
  • Probability & Statistics with Applications (AREA)
  • Chemical & Material Sciences (AREA)
  • Analytical Chemistry (AREA)
  • Biochemistry (AREA)
  • Immunology (AREA)
  • Pathology (AREA)
  • Investigating Or Analysing Materials By Optical Means (AREA)

Abstract

The invention relates to a method for tracing the origin of white tea. The method comprises the following steps: step S1, collecting white tea samples of different producing areas; s2, performing near infrared spectrum measurement on white tea samples in different producing areas to obtain white tea near infrared sample data sets in different producing areas; step S3, adding origin place identification for near-infrared white tea sample data sets of different origin places; s4, extracting characteristic data of samples in near-infrared white tea sample data sets of different producing areas by adopting a linear discriminant analysis method to obtain a characteristic data sample set; and step S5, training a multilayer perceptron model for distinguishing the white tea producing area by adopting a multilayer perceptron method, and further realizing the distinguishing of the white tea producing area. The method of the invention uses a linear discriminant analysis model and a multilayer perception model to establish near infrared spectrum prediction models of the white tea in different producing areas, and realizes the efficient and rapid prediction of the producing areas by using the near infrared spectrum information of the white tea.

Description

Method for tracing origin of white tea
Technical Field
The invention relates to a method for tracing the origin of white tea.
Background
The white tea has important economic and social values as a daily drink, and the quality of the white tea also receives more and more attention along with the improvement of the living standard of people. In order to prevent some illegal merchants from pretending to be special white tea with inferior common-grade white tea, the problem of rapidly and efficiently identifying the production place of a white tea is an urgent need to be solved.
The existing method for identifying the production place of the white tea by the naked eyes and experience of people wastes manpower, and identification precision is difficult to guarantee, so that the production place of the white tea cannot be efficiently and accurately identified.
Disclosure of Invention
The invention aims to provide a method for tracing the origin of white tea, which utilizes the frequency doubling or frequency combination vibration of chemical bonds such as C-H, N-H, O-H and the like contained in the white tea to obtain an absorption spectrum in a near infrared region in a diffuse reflection mode, and establishes near infrared spectrum prediction models of the white tea in different origins by using a linear discrimination analysis model and a multilayer perception model, thereby realizing the efficient and rapid prediction of the origin of the white tea by utilizing the near infrared spectrum information of the white tea.
In order to achieve the purpose, the technical scheme of the invention is as follows: a method for tracing the origin of white tea comprises the following steps:
step S1, collecting white tea samples of different producing areas;
s2, performing near infrared spectrum measurement on white tea samples in different producing areas to obtain white tea near infrared sample data sets in different producing areas;
step S3, adding origin place identification for near-infrared white tea sample data sets of different origin places;
s4, extracting characteristic data of samples in near-infrared white tea sample data sets of different producing areas by adopting a linear discriminant analysis method to obtain a characteristic data sample set;
and step S5, training a multilayer perceptron model for distinguishing the white tea producing area by adopting a multilayer perceptron method, and further realizing the distinguishing of the white tea producing area.
In an embodiment of the present invention, the step S1 is implemented as follows: the white tea samples of different producing areas are collected, the number of the samples is more than 100, the collection range comprises different producing enterprises in the main producing areas, and the number of the samples of the different producing enterprises is uniformly distributed.
In an embodiment of the present invention, the step S2 is implemented as follows: extracting samples from white tea samples of different producing areas, pulverizing into powder, sieving with 80 mesh sieve, and performing near infrared spectrum detection on the samples with scanning range of 4000cm-1～10000cm-1Data points at 3.857cm intervals-1The temperature range of the working environment is 15-30 ℃, and the relative humidity is less than 80%; 3 spectra were taken for each sample, and the average of the 3 spectra was taken as the final spectral plot for the corresponding sample.
In an embodiment of the present invention, the step S4 is specifically implemented as follows:
step S41, different near-infrared white tea sample data sets D ═ x1,x2,...,xmDivide into N according to the place of origin identification category labellabelsClass sample set D1,D2,...,DlabelsAnd is and
Figure BDA0002878770260000021
step S42, calculating the mean value of each sample class set:
Figure BDA0002878770260000022
wherein i represents a source identification category label, i is 1,2, … labels, and x represents a sample set D belonging to the i-th categoryiCharacteristic vector of (1), niRepresents the number of i-th samples;
total sample dataset mean:
Figure BDA0002878770260000023
wherein x represents a feature vector belonging to the sample dataset D and m represents the total number of samples;
step S43, calculating the intra-class dispersion matrix:
Figure BDA0002878770260000024
inter-class dispersion matrix:
Figure BDA0002878770260000025
step S44, linear discriminant analysis is characterized in that the more concentrated the points in the data projection categories are, the farther the points between the categories are, the better the effect is; the inter-class coupling degree is low, the intra-class polymerization degree is high, namely, the numerical value in the intra-class dispersion matrix is small, and the numerical value in the inter-class dispersion matrix is large, namely, the projection function y of the linear discriminant analysis model which can be expressed as a multi-dimensional vector is WTx,W＝(w1,w2,...,wk) Introducing Fisher criterion expression, namely a loss function:
Figure BDA0002878770260000026
when J (w) is maximum, the projection effect is optimal; optimizing the loss function, calculating
Figure BDA0002878770260000027
Finding an optimal projection matrix;
step S45, projecting the original near-infrared white tea sample data set D toBy W ═ W1,w2,...,wk) In the low-dimensional space generated for the basis vectors, the projected sample set is the feature data sample set D'.
In an embodiment of the present invention, the step S5 is specifically implemented as follows:
step S51, carrying out data segmentation on the characteristic data sample set obtained in the step S4, wherein 75% of the characteristic data sample set is used as a training set, and 25% of the characteristic data sample set is used as a testing set;
step S52, inputting the training set into a multilayer perception neural network for forward propagation, introducing an activation function into each layer of output, and except the activation function of the last layer adopting Softmax, the activation functions of other layers adopting ReLU; the activation function is formulated as:
f(Zl)＝Althe ReLU formula is:
Figure BDA0002878770260000031
the Softmax formula is:
Figure BDA0002878770260000032
wherein A islIs composed of
Figure BDA0002878770260000033
Set matrix, ZlIs composed of
Figure BDA0002878770260000034
Collecting the matrix;
the multilayer perception neural network has L layers, and the j characteristic value of the ith sample of the ith layer in forward propagation is calculated by the formula:
Figure BDA0002878770260000035
wherein, L is 1,2, … L;
step S53, calculating the error between the predicted value and the sample real label value by using a cross entropy loss function, wherein the formula is as follows:
Figure BDA0002878770260000036
wherein y is a sample true tag value;
and step S54, updating the weight coefficient of the multilayer perceptron neural network by utilizing back propagation, namely, adjusting the values of W and B to continuously reduce J to obtain a multilayer perceptron model for distinguishing the white tea producing area, thereby realizing the distinguishing of the white tea producing area.
In an embodiment of the present invention, in step S54, a derivative chain rule is adopted to reversely derive weighting coefficients of the multi-layer perceptual neural network layer by layer, and the specific implementation manner is as follows:
(1) according to a loss function J to AlDerivation to obtain dAl：
Figure BDA0002878770260000037
Figure BDA0002878770260000038
(2) According to dAlCalculating dZl：
When the activation function is Softmax, calculating dZl：
Figure BDA0002878770260000041
Figure BDA0002878770260000042
Where · is a dot product;
when the activation function is ReLU, calculating dZl：
Figure BDA0002878770260000043
Figure BDA0002878770260000044
(3) According to dZlCalculating dBl
Figure BDA0002878770260000045
So dBl＝dZl
(4) According to dBlCalculating dWl
Figure BDA0002878770260000046
Figure BDA0002878770260000047
Figure BDA0002878770260000048
(5) According to dZlCalculating dAl-1
Figure BDA0002878770260000049
Figure BDA00028787702600000410
dAl-1＝dZL·(Wl)T
(6) According to dAl-1And obtaining the update weight and the offset of all layers of the multilayer perceptive neural network by layer-by-layer recursion, wherein the formula is as follows:
Wl＝Wl-η*dWl
Bl＝Bl-η*dBl
eta is the learning rate and controls the update of the weight.
Compared with the prior art, the invention has the following beneficial effects: the method for tracing the producing area of the white tea obtains the absorption spectrum in a near infrared area in a diffuse reflection mode by utilizing the frequency doubling or frequency combination vibration of chemical bonds such as C-H, N-H, O-H and the like contained in the white tea, and establishes near infrared spectrum prediction models of the white tea in different producing areas by using a linear discriminant analysis model and a multilayer perception model, thereby realizing the efficient and rapid prediction of the producing area by utilizing the near infrared spectrum information of the white tea.
Drawings
FIG. 1 is a flowchart of a discriminant model classification method according to the present invention.
FIG. 2 is a schematic diagram of the discrimination result of the discrimination model on the white peony data set training set.
FIG. 3 is a schematic diagram of the discrimination result of the white peony data set by the discrimination model of the present invention.
Detailed Description
The technical scheme of the invention is specifically explained below with reference to the accompanying drawings.
As shown in fig. 1, the present invention provides a method for tracing the origin of white tea, comprising the following steps:
step S1, collecting white tea samples of different producing areas;
s2, performing near infrared spectrum measurement on white tea samples in different producing areas to obtain white tea near infrared sample data sets in different producing areas;
step S3, adding origin place identification for near-infrared white tea sample data sets of different origin places;
s4, extracting characteristic data of samples in near-infrared white tea sample data sets of different producing areas by adopting a linear discriminant analysis method to obtain a characteristic data sample set;
and step S5, training a multilayer perceptron model for distinguishing the white tea producing area by adopting a multilayer perceptron method, and further realizing the distinguishing of the white tea producing area.
The following is a specific implementation of the present invention.
A method for tracing the origin of white tea, comprising the following steps:
s101: collecting white tea samples of different producing areas
The number of the samples collected in different producing areas is more than 100, the collecting range includes different producing enterprises in the main producing area, and the number of the samples of the different producing enterprises is required to be uniformly distributed as much as possible.
S102: performing near infrared spectrum measurement on white tea samples of different producing areas
Taking a part of representative sample from the sample to be tested according to the method of GB/T8302, wherein the sample needs to be crushed into powder, and the granularity is required to pass through a 80-mesh sieve. Performing near infrared spectrum detection on the sample, wherein the scanning range is 4000cm-1～10000cm-1Data points at 3.857cm intervals-1The temperature range of the working environment is 15-30 ℃ and the relative humidity is less than 80% during detection. 3 spectra were taken for each sample, and the 3 spectra averaged to give the final spectral plot for that sample.
S103: augmenting source identification for sample datasets
The near infrared spectrum data of each sample is spliced into an Excel table, the wavelength of the near infrared is listed, and the near infrared data (absorbance) of each sample is obtained.
A source identification is added for each sample, at the beginning of each row, the source identification is added in the form of a number. For example, a sample from a fuding origin is identified by the numeral 1, a sample from a fuan origin is identified by the numeral 2, a sample from a political and origin is identified by the numeral 3, and a sample from a Jianyang origin is identified by the numeral 4. As shown in table 1.
TABLE 1 white peony data set
Figure BDA0002878770260000061
S104: extracting characteristic data in near infrared sample data of different producing areas
Extracting spectral characteristic data by using the sample near-infrared data set obtained in S103 and adopting a linear discriminant analysis model:
step one, a sample data set D ═ x1,x2,...,xmDivide into N according to the category labellabelsClass D1,D2,...,DlabelsAnd is and
Figure BDA0002878770260000062
Figure BDA0002878770260000063
step two, calculating the mean value (center) of each sample class set:
Figure BDA0002878770260000064
total sample dataset mean:
Figure BDA0002878770260000065
step three, calculating an intra-class dispersion matrix:
Figure BDA0002878770260000066
inter-class dispersion matrix:
Figure BDA0002878770260000067
and step four, linear discriminant analysis is characterized in that the more concentrated the points in the data projection categories are, the farther the points between the categories are, the better the effect is. The degree of coupling between the classes is low, the degree of polymerization within the classes is high, i.e., the values in the intra-class dispersion matrix are small, and the values in the inter-class dispersion matrix are large. Can be expressed as the projection function y of the multi-dimensional vector is WTx,W＝(w1,w2,...,wk) Introducing Fisher criterion expression, namely a loss function:
Figure BDA0002878770260000071
the projection effect is best when J (w) is maximum.
Step five, optimizing a loss function and calculating
Figure BDA0002878770260000072
I.e. finding the optimal projection matrix.
Step six, projecting the original sample set D to a position W ═ (W)1,w2,...,wk) In the low-dimensional space (k dimension) generated for the basis vector, k is 2 in the scheme, and the sample set after projection is the sample set D' that we need.
S105: training multilayer perceptron model
And (3) carrying out data segmentation on the sample set obtained in the S104 after the characteristics are extracted, wherein 75% of the sample set is used as a training set, and 25% of the sample set is used as a test set, and training a multilayer perceptron (MLP) model:
step one, the multilayer perceptron is a multilayer fully-connected neural network, and sample data is input into the multilayer perceptron neural network for forward propagation. And (3) introducing an activation function into the output of each layer to increase the nonlinearity of the neural network model, wherein the activation functions of other layers adopt ReLU except the activation function of the last layer adopts Softmax.
The activation function is formulated as: f (Z)l)＝AlThe ReLU formula is:
Figure BDA0002878770260000073
the Softmax formula is:
Figure BDA0002878770260000074
wherein A islIs composed of
Figure BDA0002878770260000075
Set matrix, ZlIs composed of
Figure BDA0002878770260000076
And (5) collecting the matrix.
The multilayer perception neural network has L layers, and the characteristic value of the ith sample j of the ith layer in forward propagation is calculated according to the formula:
Figure BDA0002878770260000077
step two, calculating the error between the predicted value and the sample label value by using a cross entropy loss function, wherein the formula is as follows:
Figure BDA0002878770260000078
where y is the sample true tag value.
And step three, updating the weight coefficient of the neural network by utilizing back propagation, namely continuously reducing J by adjusting the values of W and B. The layer-by-layer derivation can be reversed, based on the chain rule of the derivatives.
(1) According to a loss function J to AlDerivation to obtain dAl：
Figure BDA0002878770260000081
Figure BDA0002878770260000082
(2) According to dAlCalculating dZl：
When the activation function is Softmax, calculating dZl：
Figure BDA0002878770260000083
Figure BDA0002878770260000084
Where · is a dot product.
When the activation function is ReLU, calculating dZl：
Figure BDA0002878770260000085
Figure BDA0002878770260000086
(3) According to dZlCalculating dBl
Figure BDA0002878770260000087
So dBl＝dZl
(4) According to dBlCalculating dWl
Figure BDA0002878770260000088
Figure BDA0002878770260000089
Figure BDA00028787702600000810
(5) According to dZlCalculating dAl-1
Figure BDA0002878770260000091
Figure BDA0002878770260000092
dAl-1＝dZL·(Wl)T
(6) According to dAl-1And obtaining the update weight and the offset of all layers by layer-by-layer recursion, wherein the formula is as follows:
Wl＝Wl-η*dWl
Bl＝Bl-η*dBl
eta is the learning rate and controls the update of the weight.
And step four, repeating the steps to enable the loss function J to be as small as possible, and finally obtaining the discrimination model.
Fig. 2 and 3 are schematic diagrams of the discrimination results of the discrimination model constructed by the invention on the white peony data set training set and the white peony data set testing set.
The above are preferred embodiments of the present invention, and all changes made according to the technical scheme of the present invention that produce functional effects do not exceed the scope of the technical scheme of the present invention belong to the protection scope of the present invention.

Claims (6)

1. A method for tracing the origin of white tea is characterized by comprising the following steps:
step S1, collecting white tea samples of different producing areas;
s2, performing near infrared spectrum measurement on white tea samples in different producing areas to obtain white tea near infrared sample data sets in different producing areas;
step S3, adding origin place identification for near-infrared white tea sample data sets of different origin places;
s4, extracting characteristic data of samples in near-infrared white tea sample data sets of different producing areas by adopting a linear discriminant analysis method to obtain a characteristic data sample set;
and step S5, training a multilayer perceptron model for distinguishing the white tea producing area by adopting a multilayer perceptron method, and further realizing the distinguishing of the white tea producing area.
2. The method for tracing the production place of white tea according to claim 1, wherein said step S1 is implemented by: the white tea samples of different producing areas are collected, the number of the samples is more than 100, the collection range comprises different producing enterprises in the main producing areas, and the number of the samples of the different producing enterprises is uniformly distributed.
3. The method for tracing the production place of white tea according to claim 1, wherein said step S2 is implemented by: extracting samples from white tea samples of different producing areas, pulverizing into powder, sieving with 80 mesh sieve, and performing near infrared spectrum detection on the samples with scanning range of 4000cm-1～10000cm-1Data points at 3.857cm intervals-1The temperature range of the working environment is 15-30 ℃, and the relative humidity is less than 80%; 3 spectra were taken for each sample, and the average of the 3 spectra was taken as the final spectral plot for the corresponding sample.
4. The method for tracing the production place of white tea according to claim 1, wherein said step S4 is embodied as follows:
step S41, different near-infrared white tea sample data sets D ═ x1,x2,...,xmDivide into N according to the place of origin identification category labellabelsClass sample set D1,D2,...,DlabelsAnd D is1∪D2∪…∪Dlabels＝D，
Figure FDA0002878770250000011
Step S42, calculating the mean value of each sample class set:
Figure FDA0002878770250000012
wherein i represents a source identification category label, i is 1,2, … labels, and x represents a sample set D belonging to the i-th categoryiCharacteristic vector of (1), niRepresents the number of i-th samples;
total sample dataset mean:
Figure FDA0002878770250000013
wherein x represents a feature vector belonging to the sample dataset D and m represents the total number of samples;
step S43, calculating the intra-class dispersion matrix:
Figure FDA0002878770250000021
inter-class dispersion matrix:
Figure FDA0002878770250000022
step S44, linear discriminant analysis is characterized in that the more concentrated the points in the data projection categories are, the farther the points between the categories are, the better the effect is; the inter-class coupling degree is low, the intra-class polymerization degree is high, namely, the numerical value in the intra-class dispersion matrix is small, and the numerical value in the inter-class dispersion matrix is large, namely, the projection function y of the linear discriminant analysis model which can be expressed as a multi-dimensional vector is WTx,W＝(w1,w2,...,wk) Introducing Fisher criterion expression, namely a loss function:
Figure FDA0002878770250000023
when J (w) is maximum, the projection effect is optimal; optimizing the loss function, calculating
Figure FDA0002878770250000024
Finding an optimal projection matrix;
step S45, projecting the original near-infrared white tea sample dataset D onto a sample set of white tea samples W ═ W (W ═ W)1,w2,...,wk) In the low-dimensional space generated for the basis vectors, the projected sample set is the feature data sample set D'.
5. The method for tracing the production place of white tea according to claim 1, wherein said step S5 is embodied as follows:
step S51, carrying out data segmentation on the characteristic data sample set obtained in the step S4, wherein 75% of the characteristic data sample set is used as a training set, and 25% of the characteristic data sample set is used as a testing set;
step S52, inputting the training set into a multilayer perception neural network for forward propagation, introducing an activation function into each layer of output, and except the activation function of the last layer adopting Softmax, the activation functions of other layers adopting ReLU; the activation function is formulated as:
f(Zl)＝Althe ReLU formula is:
Figure FDA0002878770250000025
the Softmax formula is:
Figure FDA0002878770250000026
wherein A islIs composed of
Figure FDA0002878770250000027
Set matrix, ZlIs composed of
Figure FDA0002878770250000028
Collecting the matrix;
the multilayer perception neural network has L layers, and the j characteristic value of the ith sample of the ith layer in forward propagation is calculated by the formula:
Figure FDA0002878770250000029
wherein, L is 1,2, … L;
step S53, calculating the error between the predicted value and the sample real label value by using a cross entropy loss function, wherein the formula is as follows:
Figure FDA0002878770250000031
wherein y is a sample true tag value;
and step S54, updating the weight coefficient of the multilayer perceptron neural network by utilizing back propagation, namely, adjusting the values of W and B to continuously reduce J to obtain a multilayer perceptron model for distinguishing the white tea producing area, thereby realizing the distinguishing of the white tea producing area.
6. The method for tracing the production place of white tea as claimed in claim 1, wherein in step S54, the weight coefficients of the multi-layer perceptual neural network are reversely derived layer by using a derivative chain rule, and the specific implementation manner is as follows:
(1) according to a loss function J to AlDerivation to obtain dAl：
Figure FDA0002878770250000032
Figure FDA0002878770250000033
(2) According to dAlCalculating dZl：
When the activation function is Softmax, calculating dZl：
Figure FDA0002878770250000034
Figure FDA0002878770250000035
Where · is a dot product;
when the activation function is ReLU, calculating dZl：
Figure FDA0002878770250000036
Figure FDA0002878770250000037
(3) According to dZlCalculating dBl
Figure FDA0002878770250000041
So dBl＝dZl
(4) According to dBlCalculating dWl
Figure FDA0002878770250000042
Figure FDA0002878770250000043
Figure FDA0002878770250000044
(5) According to dZlCalculating dAl-1
Figure FDA0002878770250000045
Figure FDA0002878770250000046
dAl-1＝dZL·(Wl)T
(6) According to dAl-1And obtaining the update weight and the offset of all layers of the multilayer perceptive neural network by layer-by-layer recursion, wherein the formula is as follows:
Wl＝Wl-η*dWl
Bl＝Bl-η*dBl
eta is the learning rate and controls the update of the weight.
CN202011643531.4A 2020-12-31 2020-12-31 Method for tracing origin of white tea Pending CN112861907A (en)

Priority Applications (1)

Application Number Priority Date Filing Date Title
CN202011643531.4A CN112861907A (en) 2020-12-31 2020-12-31 Method for tracing origin of white tea

Applications Claiming Priority (1)

Application Number Priority Date Filing Date Title
CN202011643531.4A CN112861907A (en) 2020-12-31 2020-12-31 Method for tracing origin of white tea

Publications (1)

Publication Number Publication Date
CN112861907A true CN112861907A (en) 2021-05-28

Family

ID=76000897

Family Applications (1)

Application Number Title Priority Date Filing Date
CN202011643531.4A Pending CN112861907A (en) 2020-12-31 2020-12-31 Method for tracing origin of white tea

Country Status (1)

Country Link
CN (1) CN112861907A (en)

Cited By (1)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN114295580A (en) * 2021-12-29 2022-04-08 四川启睿克科技有限公司 Method for rapidly judging pepper quality based on near infrared spectrum

Citations (9)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN105938093A (en) * 2016-06-08 2016-09-14 福建农林大学 Oolong tea producing area discrimination method based on combination of genetic algorithm and support vector machine
CN106560697A (en) * 2016-10-20 2017-04-12 中国计量大学 Method for identifying producing area of Wuyi rock tea through combination of near infrared spectroscopy and trace element detection
CN107037039A (en) * 2017-04-19 2017-08-11 塔里木大学 A kind of Xinjiang walnut place of production is traced to the source research method
CN107273926A (en) * 2017-06-12 2017-10-20 大连海事大学 A kind of linear discriminant analysis dimension reduction method weighted based on cosine similarity
CN109191564A (en) * 2018-07-27 2019-01-11 中国科学院自动化研究所 Exciting tomography fluorescence imaging three-dimensional rebuilding method based on deep learning
CN110308109A (en) * 2019-05-05 2019-10-08 湖北省农业科学院农业质量标准与检测技术研究所 A method of perchlorate content range in tealeaves is differentiated based near infrared spectrum
CN110717368A (en) * 2018-07-13 2020-01-21 北京服装学院 Qualitative classification method for textiles
CN111027627A (en) * 2019-12-11 2020-04-17 哈尔滨高斯触控科技有限公司 Vibration information terrain classification and identification method based on multilayer perceptron
CN111896495A (en) * 2020-08-05 2020-11-06 安徽大学 Method and system for discriminating Taiping Houkui production places based on deep learning and near infrared spectrum

Patent Citations (9)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN105938093A (en) * 2016-06-08 2016-09-14 福建农林大学 Oolong tea producing area discrimination method based on combination of genetic algorithm and support vector machine
CN106560697A (en) * 2016-10-20 2017-04-12 中国计量大学 Method for identifying producing area of Wuyi rock tea through combination of near infrared spectroscopy and trace element detection
CN107037039A (en) * 2017-04-19 2017-08-11 塔里木大学 A kind of Xinjiang walnut place of production is traced to the source research method
CN107273926A (en) * 2017-06-12 2017-10-20 大连海事大学 A kind of linear discriminant analysis dimension reduction method weighted based on cosine similarity
CN110717368A (en) * 2018-07-13 2020-01-21 北京服装学院 Qualitative classification method for textiles
CN109191564A (en) * 2018-07-27 2019-01-11 中国科学院自动化研究所 Exciting tomography fluorescence imaging three-dimensional rebuilding method based on deep learning
CN110308109A (en) * 2019-05-05 2019-10-08 湖北省农业科学院农业质量标准与检测技术研究所 A method of perchlorate content range in tealeaves is differentiated based near infrared spectrum
CN111027627A (en) * 2019-12-11 2020-04-17 哈尔滨高斯触控科技有限公司 Vibration information terrain classification and identification method based on multilayer perceptron
CN111896495A (en) * 2020-08-05 2020-11-06 安徽大学 Method and system for discriminating Taiping Houkui production places based on deep learning and near infrared spectrum

Non-Patent Citations (2)

* Cited by examiner, † Cited by third party
Title
徐鹏程: "龙井茶HPLC品质指纹图谱的构建与应用研究", 《中国优秀博硕士学位论文全文数据库(硕士)工程科技Ⅰ辑》 *
袁远: "基于数据挖掘的绿色设计中客户需求向工程特性转化研究", 《中国优秀博硕士学位论文全文数据库(硕士)工程科技Ⅱ辑》 *

Cited By (2)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN114295580A (en) * 2021-12-29 2022-04-08 四川启睿克科技有限公司 Method for rapidly judging pepper quality based on near infrared spectrum
CN114295580B (en) * 2021-12-29 2023-07-11 四川启睿克科技有限公司 Method for rapidly judging quality of peppers based on near infrared spectrum

Similar Documents

Publication Publication Date Title
CN106599797B (en) A kind of infrared face recognition method based on local parallel neural network
CN105469034B (en) Face identification method based on Weighting type distinctive sparse constraint Non-negative Matrix Factorization
CN106295694B (en) Face recognition method for iterative re-constrained group sparse representation classification
CN109632693A (en) A kind of tera-hertz spectra recognition methods based on BLSTM-RNN
CN109102029B (en) Method for evaluating quality of synthesized face sample by using information maximization generation confrontation network model
CN104866831B (en) The face recognition algorithms of characteristic weighing
CN109815892A (en) The signal recognition method of distributed fiber grating sensing network based on CNN
CN113869162A (en) Violation identification method and system based on artificial intelligence
CN109190698B (en) Classification and identification system and method for network digital virtual assets
CN110490659B (en) GAN-based user load curve generation method
CN101599126A (en) Utilize the support vector machine classifier of overall intercommunication weighting
CN107203752A (en) A kind of combined depth study and the face identification method of the norm constraint of feature two
CN110321777A (en) A kind of face identification method based on the sparse denoising self-encoding encoder of stack convolution
CN109919921B (en) Environmental impact degree modeling method based on generation countermeasure network
CN109657707A (en) A kind of image classification method based on observing matrix transformation dimension
CN112347910A (en) Signal fingerprint identification method based on multi-mode deep learning
CN112861907A (en) Method for tracing origin of white tea
CN114298834A (en) Personal credit evaluation method and system based on self-organizing mapping network
Zhang Application of artificial intelligence recognition technology in digital image processing
CN110458071B (en) DWT-DFPA-GBDT-based optical fiber vibration signal feature extraction and classification method
CN112966735B (en) Method for fusing supervision multi-set related features based on spectrum reconstruction
CN117422936A (en) Remote sensing image classification method and system
CN111950429A (en) Face recognition method based on weighted collaborative representation
CN117113243A (en) Photovoltaic equipment abnormality detection method
CN116885697A (en) Load prediction method based on combination of cluster analysis and intelligent algorithm

Legal Events

Date Code Title Description
PB01 Publication
PB01 Publication
SE01 Entry into force of request for substantive examination
SE01 Entry into force of request for substantive examination
RJ01 Rejection of invention patent application after publication

Application publication date: 20210528

RJ01 Rejection of invention patent application after publication