CN110308111B - Method for rapidly predicting time for smoldering yellow tea by using near infrared spectrum technology - Google Patents
Method for rapidly predicting time for smoldering yellow tea by using near infrared spectrum technology Download PDFInfo
- Publication number
- CN110308111B CN110308111B CN201910513186.3A CN201910513186A CN110308111B CN 110308111 B CN110308111 B CN 110308111B CN 201910513186 A CN201910513186 A CN 201910513186A CN 110308111 B CN110308111 B CN 110308111B
- Authority
- CN
- China
- Prior art keywords
- spectrum
- sample
- model
- yellow
- samples
- 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.)
- Active
Links
- 238000000034 method Methods 0.000 title claims abstract description 73
- 235000020338 yellow tea Nutrition 0.000 title claims abstract description 38
- 238000002329 infrared spectrum Methods 0.000 title claims abstract description 31
- 238000005516 engineering process Methods 0.000 title claims abstract description 12
- 238000001228 spectrum Methods 0.000 claims abstract description 96
- 238000013528 artificial neural network Methods 0.000 claims abstract description 54
- 230000003595 spectral effect Effects 0.000 claims abstract description 42
- 210000002569 neuron Anatomy 0.000 claims abstract description 31
- 238000012546 transfer Methods 0.000 claims abstract description 31
- 238000004383 yellowing Methods 0.000 claims abstract description 29
- 238000012843 least square support vector machine Methods 0.000 claims abstract description 27
- 238000007781 pre-processing Methods 0.000 claims abstract description 13
- 238000012360 testing method Methods 0.000 claims abstract description 9
- 238000012847 principal component analysis method Methods 0.000 claims abstract description 5
- 238000012937 correction Methods 0.000 claims description 28
- 238000000513 principal component analysis Methods 0.000 claims description 26
- 238000012795 verification Methods 0.000 claims description 21
- 230000000694 effects Effects 0.000 claims description 15
- 238000004364 calculation method Methods 0.000 claims description 12
- 238000002790 cross-validation Methods 0.000 claims description 9
- 238000012216 screening Methods 0.000 claims description 8
- 238000010606 normalization Methods 0.000 claims description 7
- 102000004190 Enzymes Human genes 0.000 claims description 5
- 108090000790 Enzymes Proteins 0.000 claims description 5
- 229910000530 Gallium indium arsenide Inorganic materials 0.000 claims description 4
- 235000020334 white tea Nutrition 0.000 claims description 4
- 230000001105 regulatory effect Effects 0.000 claims description 2
- 230000002708 enhancing effect Effects 0.000 abstract description 3
- 241001122767 Theaceae Species 0.000 description 12
- 235000013616 tea Nutrition 0.000 description 12
- 238000012545 processing Methods 0.000 description 7
- 238000010200 validation analysis Methods 0.000 description 7
- 238000012706 support-vector machine Methods 0.000 description 5
- 239000010410 layer Substances 0.000 description 4
- 235000014347 soups Nutrition 0.000 description 4
- 230000009286 beneficial effect Effects 0.000 description 3
- ATNHDLDRLWWWCB-AENOIHSZSA-M chlorophyll a Chemical compound C1([C@@H](C(=O)OC)C(=O)C2=C3C)=C2N2C3=CC(C(CC)=C3C)=[N+]4C3=CC3=C(C=C)C(C)=C5N3[Mg-2]42[N+]2=C1[C@@H](CCC(=O)OC\C=C(/C)CCC[C@H](C)CCC[C@H](C)CCCC(C)C)[C@H](C)C2=C5 ATNHDLDRLWWWCB-AENOIHSZSA-M 0.000 description 2
- 238000007405 data analysis Methods 0.000 description 2
- 230000002349 favourable effect Effects 0.000 description 2
- 238000009499 grossing Methods 0.000 description 2
- FFRBMBIXVSCUFS-UHFFFAOYSA-N 2,4-dinitro-1-naphthol Chemical compound C1=CC=C2C(O)=C([N+]([O-])=O)C=C([N+]([O-])=O)C2=C1 FFRBMBIXVSCUFS-UHFFFAOYSA-N 0.000 description 1
- 235000009024 Ceanothus sanguineus Nutrition 0.000 description 1
- 240000003553 Leptospermum scoparium Species 0.000 description 1
- 235000015459 Lycium barbarum Nutrition 0.000 description 1
- 244000018633 Prunus armeniaca Species 0.000 description 1
- 235000009827 Prunus armeniaca Nutrition 0.000 description 1
- 230000002411 adverse Effects 0.000 description 1
- 230000032683 aging Effects 0.000 description 1
- 150000001413 amino acids Chemical class 0.000 description 1
- 229930002875 chlorophyll Natural products 0.000 description 1
- 235000019804 chlorophyll Nutrition 0.000 description 1
- 229930002868 chlorophyll a Natural products 0.000 description 1
- 229930002869 chlorophyll b Natural products 0.000 description 1
- NSMUHPMZFPKNMZ-VBYMZDBQSA-M chlorophyll b Chemical compound C1([C@@H](C(=O)OC)C(=O)C2=C3C)=C2N2C3=CC(C(CC)=C3C=O)=[N+]4C3=CC3=C(C=C)C(C)=C5N3[Mg-2]42[N+]2=C1[C@@H](CCC(=O)OC\C=C(/C)CCC[C@H](C)CCC[C@H](C)CCCC(C)C)[C@H](C)C2=C5 NSMUHPMZFPKNMZ-VBYMZDBQSA-M 0.000 description 1
- 230000003749 cleanliness Effects 0.000 description 1
- 235000009508 confectionery Nutrition 0.000 description 1
- 238000010276 construction Methods 0.000 description 1
- 230000001276 controlling effect Effects 0.000 description 1
- 230000001186 cumulative effect Effects 0.000 description 1
- 230000009849 deactivation Effects 0.000 description 1
- 230000007547 defect Effects 0.000 description 1
- 238000010586 diagram Methods 0.000 description 1
- 238000001035 drying Methods 0.000 description 1
- 239000003205 fragrance Substances 0.000 description 1
- 230000010365 information processing Effects 0.000 description 1
- 230000002452 interceptive effect Effects 0.000 description 1
- 239000011229 interlayer Substances 0.000 description 1
- 230000002045 lasting effect Effects 0.000 description 1
- 238000012821 model calculation Methods 0.000 description 1
- 238000003062 neural network model Methods 0.000 description 1
- 230000003287 optical effect Effects 0.000 description 1
- 150000008442 polyphenolic compounds Chemical class 0.000 description 1
- 235000013824 polyphenols Nutrition 0.000 description 1
- 102000004169 proteins and genes Human genes 0.000 description 1
- 108090000623 proteins and genes Proteins 0.000 description 1
- 238000011160 research Methods 0.000 description 1
- 239000002689 soil Substances 0.000 description 1
- 238000000638 solvent extraction Methods 0.000 description 1
- 238000006467 substitution reaction Methods 0.000 description 1
- 238000013519 translation Methods 0.000 description 1
- XLYOFNOQVPJJNP-UHFFFAOYSA-N water Substances O XLYOFNOQVPJJNP-UHFFFAOYSA-N 0.000 description 1
Images
Classifications
-
- 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
- G06F18/21—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
- G06F18/2135—Feature extraction, e.g. by transforming the feature space; Summarisation; Mappings, e.g. subspace methods based on approximation criteria, e.g. principal component analysis
-
- 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
- G06F18/2411—Classification techniques relating to the classification model, e.g. parametric or non-parametric approaches based on the proximity to a decision surface, e.g. support vector machines
-
- 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
- G01N2021/3595—Investigating relative effect of material at wavelengths characteristic of specific elements or molecules, e.g. atomic absorption spectrometry using infrared light using FTIR
Landscapes
- Engineering & Computer Science (AREA)
- Physics & Mathematics (AREA)
- Data Mining & Analysis (AREA)
- Theoretical Computer Science (AREA)
- General Physics & Mathematics (AREA)
- Life Sciences & Earth Sciences (AREA)
- Computer Vision & Pattern Recognition (AREA)
- Evolutionary Computation (AREA)
- Evolutionary Biology (AREA)
- Bioinformatics & Computational Biology (AREA)
- Bioinformatics & Cheminformatics (AREA)
- Artificial Intelligence (AREA)
- General Engineering & Computer Science (AREA)
- Spectroscopy & Molecular Physics (AREA)
- Analytical Chemistry (AREA)
- Pathology (AREA)
- Immunology (AREA)
- General Health & Medical Sciences (AREA)
- Biochemistry (AREA)
- Chemical & Material Sciences (AREA)
- Health & Medical Sciences (AREA)
- Investigating Or Analysing Materials By Optical Means (AREA)
Abstract
A method for rapidly predicting the time of smoldering yellow of Yuanan yellow tea by applying a near infrared spectrum technology comprises the following steps: collecting and classifying fresh leaf samples; scanning to obtain near infrared spectrums of fresh leaf samples at different yellowing time; preprocessing a sample spectrum to remove noise information, and converting the sample spectrum into paired data points; all the spectral data are divided into 20 subintervals, a least square support vector machine method model of each subinterval data is respectively established, and the best subinterval data for modeling are screened out; extracting and compressing the optimal spectrum subinterval information by using a principal component analysis method; continuously adjusting the number of neurons and a transfer function by taking the principal component score as an input value, and establishing an unsupervised Kohonen structure artificial neural network prediction model; and (5) testing the robustness of the model. The fast, accurate and objective prediction of the yellow smoldering time of the yellow tea sample is realized, and the purposes of improving the accuracy of the predicted yellow smoldering time and enhancing the practicability of a model are achieved.
Description
Technical Field
The invention relates to a method for predicting yellow tea smoldering time, in particular to a method for rapidly predicting far-safe yellow tea smoldering time by applying a near infrared spectrum technology.
Background
Yellow tea is one of six major tea types in China, and the Hubei province has the longest history of far-safe yellow tea. The far-distance safe yellow tea is honorable in climate in Hubei, abundant in rainfall, loose and fertile in soil, and favorable in ecological environment, is very favorable for the growth of tea trees, and ensures the excellent quality of tea leaves, so that the far-distance safe yellow tea is praised as a good product in the Hubei tea, is a tea product which is well sold in the tea market, is not supplied and demanded every year, and is deeply favored by people.
During processing, the fresh leaves of the far-safety yellow tea are generally picked according to the standard of single bud, one bud and one leaf and two leaves, the fresh leaves are required to be tender and fresh, the cleanliness of the fresh leaves is good, the processed yellow tea is golden yellow in color and exposed in pekoe, the fragrance is fragrant and lasting, the taste is mellow and sweet, the liquor color is apricot yellow and bright, and the leaf bottom is tender, yellow and uniform. The basic processing technology of the far-safety yellow tea comprises the following steps: deactivating enzyme, closing to yellow and drying; the yellow-smoldering process is the most key processing process and is the basis for forming the unique 'golden dry tea, yellow and bright soup color and tender yellow leaf bottom' of the yellow tea. The yellow tea smoldering time and the content components have close correspondence, and researches such as aging indicate that: with the increase of the yellow-smoldering time, the color of the dry tea is reduced, the yellow color is exposed, the color of the tea soup is changed from green to bright, the taste is fresh, mellow and refreshing, and the tea soup is slightly tender and fragrant; in the process of yellow-stewing, the content of polyphenols is reduced, the content of amino acid is increased, the content of soluble protein is slowly reduced, the content of soluble sugar is slightly increased, the content of water extract is slightly increased along with the prolonging of yellow-stewing time, and the content of chlorophyll a, chlorophyll b and total chlorophyll is gradually reduced. Therefore, under the same conditions, the yellow stewing time is a key factor influencing the taste and quality of the yellow tea soup. Therefore, timely and accurately controlling the yellow tea smoldering time is beneficial to improving the quality of the yellow tea, improving the commodity value and improving the economic value of the yellow tea; and the method is also beneficial to local tea farmers to enhance the self economic strength, improve the self living conditions and realize the goal of early delightful as soon as possible. Therefore, how to judge the yellow-smoldering time of the yellow tea in real time is very important.
At present, the far-safety yellow tea is usually subjected to manual recording of the yellow smoldering time by self, but in the busy season of yellow tea processing, due to the fact that labor intensity is high, tea processing personnel are prone to over-fatigue, the manual timing method is high in subjectivity and cannot accurately master the yellow smoldering time, the situation that yellow tea is not fully yellow smoldered or is excessively yellow smoldered is prone to occurring, the quality of the yellow tea is possibly reduced, and great economic loss is brought to processing factories and dealers. Therefore, a method for predicting the time for smoldering yellow tea in a long-distance safe manner in a timely, accurate and objective manner is urgently needed.
Disclosure of Invention
The invention aims to provide a method for rapidly predicting the smoldering yellow time of far-safety yellow tea by applying a near infrared spectrum technology, aiming at the defects that the smoldering yellow time cannot be accurately mastered in real time by manually recording the smoldering yellow time, the tea quality is easily reduced and the like of the existing far-safety yellow tea.
In order to achieve the purpose, the technical solution of the invention is as follows: a method for rapidly predicting the time of smoldering yellow of Yuan' an yellow tea by applying a near infrared spectrum technology comprises the steps of scanning to obtain near infrared spectrums of fresh leaf samples at different smoldering yellow times, preprocessing the spectrums of the samples to remove noise information, and converting the spectrums of the samples into paired data points; all the spectral data are divided into 20 subintervals, a least square support vector machine method model of each subinterval data is respectively established, and the best subinterval data for modeling are screened out; and then compressing and extracting the optimal subinterval information data points by using a principal component analysis method, establishing artificial neural network prediction models of different yellow-smoldering time samples by using the principal component score as an input numerical value and regulating the number of neurons in a hidden layer and an information transfer function, wherein the artificial neural network prediction models are used for predicting the yellow-smoldering time of the yellow tea, and the method specifically comprises the following steps:
step one, fresh leaf sample collection and classification
Collecting three fresh leaf samples of single bud, one bud and one leaf and one bud and two leaves at different positions of an Anji white tea variety in Yuan-an county of Hubei province, deactivating enzymes, performing smoldering yellowing on the samples, and simultaneously accurately recording smoldering yellowing time; according to different dark yellow time, dividing the sample into 2 sets of a correction set and a verification set, and respectively establishing a correction set near infrared spectrum prediction model and testing the robustness of the correction set prediction model;
wherein, one bud and one leaf are composed of single bud, first leaf and long stalk, and one bud and two leaves are composed of single bud, first leaf, second leaf and long stalk;
step two, spectrum scanning
Scanning by a near-infrared spectrometer to obtain the near-infrared spectrum of all the smoldering yellow samples, wherein the spectrum scanning range is 4000-10000cm-1Resolution of 8cm-1The detector is InGaAs, 3 spectra are collected by each sample, each time of scanning is carried out for 64 times, then the 3 collected spectra are averaged, and the average spectrum is used as the final spectrum of the sample for subsequent modeling;
thirdly, preprocessing the spectral noise information
Performing denoising pretreatment on the near infrared spectrum obtained in the step two by using a vector normalization method by using multiple chemometrics software, so that the signal-to-noise ratio of the spectrum is improved, and a stable prediction model is favorably established; after the spectrum is denoised, converting the sample spectrum into paired data points;
step four, spectrum subinterval division
Dividing all spectral information data into 20 spectral information subintervals equally, and accurately screening the spectral information subintervals reflecting the sample yellowing time for subsequently establishing a least square support vector machine model;
step five, establishing a least square support vector machine (LS-SVM) model
The invention respectively establishes a prediction model of each spectrum information subinterval by applying an LS-SVM method, compares the correlation coefficient of the model (Rc) with the magnitude of cross validation root mean square variance (RMSECV), preliminarily screens out the best spectrum information subinterval data of the model, achieves the purpose of accurately screening the spectrum information reflecting the yellow-smoldering time, wherein the Rc is the largest, the RMSECV is the smallest, and represents that the result of the established least square support vector machine model is the best,
wherein, the RMSECV calculation formula is as follows:
the Rc is calculated by the formula:
wherein n represents the number of samples, yi and yi' measured value and predicted value of the ith sample in the sample set respectively,
the average value of the measured values of the ith sample in the sample set is shown, wherein i is less than or equal to n;
step six, establishing unsupervised Kohonen structure artificial neural network prediction model
Further accurately predicting the yellowing time of the sample by using a nonlinear artificial neural network method, wherein the method comprises the following steps:
1) principal component analysis of optimal spectral information subintervals
Performing principal component analysis on the screened optimal spectrum information subinterval data by adopting a Principal Component Analysis (PCA) method to obtain an individual contribution rate value, an accumulated contribution rate value and a principal component score of each principal component; the sample modeling spectrum subinterval information can be effectively represented only when the accumulated contribution rate of the principal components is more than or equal to 85%;
2) artificial neural network prediction model establishment
Establishing an artificial neural network prediction model of an unsupervised Kohonen structure by using main component scores of an optimal spectral information subinterval as an input value and different yellowing times of a sample as an output value and applying Neuro Shell 2 software, wherein the unsupervised Kohonen structure artificial neural network comprises three information transfer functions of 1 hidden layer, 3, 4 and 5 neuron numbers, logic, line [0,1] and tanh, and comparing a correlation coefficient (Rc) of the model with the size of a cross validation Root Mean Square Error (RMSECV) to obtain the optimal near infrared spectrum prediction model, wherein the larger the Rc is, the smaller the RMSECV is, the better the model prediction effect is represented;
wherein the RMSECV calculation formula is as follows:
the Rc is calculated by the formula:
wherein n represents the number of samples, yi and yi' measured value and predicted value of the ith sample in the sample set respectively,
the average value of the measured values of the ith sample in the sample set is shown, wherein i is less than or equal to n;
the model with the maximum correlation coefficient Rc and the minimum cross validation root mean square deviation RMSECV is used as an optimal model, and an optimal correction set model is obtained after comparison;
step seven, testing the robustness of the model
The artificial neural network prediction model effects of different yellowing times are tested by applying all verification set samples, and the obtained result is represented by a correlation coefficient (Rp) and a verification mean square error (RMSEP), wherein the larger the Rp and the smaller the RMSEP, the more the model robustness is represented, and the yellowing time of the sample can be accurately predicted;
wherein the RMSEP calculation formula is as follows:
the calculation formula of Rp is as follows:
wherein n represents the number of samples, yi and yi' the measured value and the predicted value of the ith sample in the sample set are respectively, wherein i is less than or equal to n.
In the first step, the number of the fresh leaf samples is 120, the fresh leaf samples are divided into a correction set and a verification set according to the ratio of 3:1, wherein 90 correction set samples and 30 verification set samples are obtained.
The sub-interval data of the optimal spectrum information screened and modeled in the step five is 6406.4-6703.4cm-1。
And sixthly, when Principal Component Analysis (PCA) is adopted to carry out principal component analysis on the screened optimal spectrum information subinterval data, the first 3 principal components are adopted to represent the spectrum information of the optimal spectrum subinterval.
And in the sixth step, 4 neurons and logistic information transfer functions are selected to establish unsupervised Kohonen structure artificial neural network models of different samples with different dark yellow time.
Compared with the prior art, the invention has the beneficial effects that:
1. according to the method, after the noise information of a yellow sample is removed, the spectrum of the sample is converted into paired data points for storage, all spectrum data are divided into 20 subintervals, a least square support vector machine method model of each subinterval data is established respectively, and the best subinterval data for modeling is screened out; the principal component analysis method is applied to carry out principal component analysis on the optimal subinterval data points, and spectral information is compressed and extracted, so that the calculation amount of the model is reduced, and the robustness of the model is improved; and then, establishing artificial neural network prediction models of different yellowing times by taking the principal component score as an input value, realizing the rapid, accurate and objective prediction of the yellowing time of the sample, and achieving the purposes of improving the accuracy of the prediction of the yellowing time and enhancing the practicability of the models.
2. The method accurately screens the spectral information reflecting the sample yellowing time by using a least square support vector machine method, obtains the principal component score of the optimal spectral information by continuously practicing comparison and prediction effects, takes the principal component score as input data, and achieves the purpose of accurately predicting the sample yellowing time by continuously and repeatedly optimizing the number of neurons and transfer functions in the unsupervised Kohonen structure artificial neural network method.
Drawings
FIG. 1 is a graph of the different yellowing time near infrared spectra of all 120 fresh leaf samples in the present invention.
FIG. 2 is a schematic diagram of the unsupervised Kohonen neural network model structure of the present invention.
Detailed Description
The invention is described in further detail below with reference to the following description of the drawings and the detailed description.
Referring to fig. 1 to fig. 2, a method for rapidly predicting the yellow smoldering time of the far-safe yellow tea by using a near infrared spectrum technology, scanning to obtain near infrared spectrums of fresh leaf samples at different yellow smoldering times, preprocessing the sample spectrum to remove noise information, and converting the sample spectrum into paired data points to be stored in an excel table; then dividing all the spectral data into 20 subintervals, respectively establishing a least square support vector machine method model of each subinterval data, and screening out the best subinterval data for modeling; and then compressing and extracting subinterval information data points by using a principal component analysis method, taking principal component scores as input numerical values, and establishing artificial neural network prediction models of different yellow-smoldering time samples by adjusting the number of neurons in the hidden layer and an information transfer function, wherein the artificial neural network prediction models are used for predicting the yellow-smoldering time of the yellow tea. The method specifically comprises the following steps:
step one, fresh leaf sample collection and classification
Collecting three fresh leaf samples of single bud, one bud and one leaf and one bud and two leaves at different positions of an Anji white tea variety in Yuan-an county of Hubei province, deactivating enzymes, performing smoldering yellowing on the samples, and simultaneously accurately recording smoldering yellowing time; according to different dark yellow time, the sample is divided into 2 sets of a correction set and a verification set, and the correction set near infrared spectrum prediction model are respectively established and tested for robustness.
Wherein, one bud and one leaf on the upper surface are composed of a single bud, a first leaf and a long stalk, and the two leaves on the upper surface are composed of a single bud, a first leaf, a second leaf and a long stalk;
step two, spectrum scanning
Scanning by using an America Saimei-Antaris II type Fourier transform near-infrared spectrometer to obtain the near-infrared spectrum of all the smoldering yellow samples, wherein the spectrum scanning range is 4000-10000cm-1Resolution of 8cm-1The detector is InGaAs, each sample acquires 3 spectra and scans 64 times each time; the 3 acquired spectra were then averaged and the average spectrum was used as the final spectrum for the sample to be subsequently modeled.
Thirdly, preprocessing the spectral noise information
Because background information of high-frequency noise and base line disturbance exists in the spectrum scanning process, if the spectrum noise is not preprocessed and is directly used for building a prediction model, the prediction effect of the model is poor, and the model is unstable, so that the original spectrum information is subjected to denoising processing before modeling. In the step, a plurality of chemometrics software is applied to perform denoising pretreatment on the near infrared spectrum obtained in the step two by adopting a smoothing method, a first derivative method, a second derivative method, a multivariate scattering correction method and a vector normalization method, so that the signal-to-noise ratio of the spectrum is improved, and a stable prediction model is favorably established; the vector normalization preprocessing method can deduct the influence of linear translation in the sample spectrum, and can independently correct each spectrum, has strong information processing capacity, and is an optimal spectrum preprocessing method.
After the spectrum is denoised, the sample spectrum is converted into paired data points (X-Y one-to-one correspondence) and stored in an excel table.
Step four, spectrum subinterval division
The near infrared spectrum contains all information of the sample, such as producing area, picking time, yellow smoldering time, content component information and the like, so that in order to improve the prediction effect of the model, a spectrum information wave band reflecting the yellow smoldering time of the sample needs to be screened, and useless spectrum information is removed and modeled; the method can improve the model prediction accuracy, greatly reduce the computation of the model, reduce the computation time of modeling and reduce the modeling cost. Therefore, the method equally divides all spectral information data into 20 spectral information subintervals, and accurately screens the spectral information subintervals reflecting the sample dark yellow time for subsequently establishing the least square support vector machine model.
Step five, establishing a least square support vector machine (LS-SVM) model
The least square support vector machine (LS-SVM) model is a classification method based on a statistical theory, mainly realizes the prediction of the model by constructing a separating hyperplane, and is widely applied with good generalization capability and robustness; meanwhile, the LS-SVM method realizes a final decision function by solving a linear equation set, reduces the solving difficulty to a certain extent, and improves the solving speed, so that the LS-SVM method is more suitable for general practical application. Therefore, in order to better predict the yellow smoldering time of the far-safety yellow tea sample, the method respectively establishes a prediction model of each spectral information subinterval by applying an LS-SVM method, compares the correlation coefficient (Rc) of the model with the cross validation Root Mean Square Error (RMSECV) of the model, preliminarily screens out the best spectral information subinterval data of the model, and achieves the purpose of accurately screening the spectral information reflecting the yellow smoldering time. Wherein Rc is the largest, RMSECV is the smallest, and the result of the established least square support vector machine model is the best;
wherein, the RMSECV calculation formula is as follows:
the Rc is calculated by the formula:
wherein n represents the number of samples, yi and yi' measured value and predicted value of the ith sample in the sample set respectively,
is the average value of the measured values of the ith sample in the sample set, wherein i is less than or equal to n.
Meanwhile, the LS-SVM method also verifies which optimal spectrum preprocessing method is selected in the third step.
Step six, establishing unsupervised Kohonen structure artificial neural network prediction model
On the basis of the fifth step, although spectrum information reflecting the stuffy yellow time is obtained preliminarily, a plurality of spectrum data points are input into the model, and a nonlinear relation is likely to exist among the data points, so that the stuffy yellow time of the sample is predicted more accurately by applying a nonlinear artificial neural network method. The method comprises the following steps:
1) principal component analysis of optimal spectral information subintervals
When the artificial neural network method is established, less input data is required, so that the spectral information of the sample needs to be further compressed and extracted, and the Principal Component Analysis (PCA) is an effective method for extracting the spectral information. Therefore, Principal Component Analysis (PCA) is performed on the screened optimal spectrum information subinterval data by applying a principal component analysis program in Matlab 2012a software to obtain an individual contribution rate value, an accumulated contribution rate value and a principal component score of each principal component, and the sample modeling spectrum subinterval information can be effectively represented only when the accumulated contribution rate of the principal components is greater than or equal to 85%.
2) Artificial neural network prediction model establishment
Establishing an artificial neural network prediction model of an unsupervised Kohonen structure by applying Neuro Shell 2 software by taking the principal component score of the optimal spectrum information subinterval as an input value and different yellowing time of the sample as an output value; the unsupervised Kohonen structure artificial neural network method contains three information transfer functions of 1 hidden layer, 3, 4 and 5 neuron numbers and logistic, linear [0,1] and tanh. In order to achieve the best prediction effect, a large amount of experimental data is needed to further screen the optimal combination of the number of neurons and the transfer function of the obtained 9 unsupervised Kohonen structure artificial neural networks, so that the best prediction effect can be achieved; therefore, the optimal near infrared spectrum prediction model is obtained by comparing the correlation coefficient (Rc) of the model with the magnitude of the Root Mean Square Error (RMSECV) of the cross validation, wherein the larger the Rc is, the smaller the RMSECV is, the better the model prediction effect is.
Wherein, the RMSECV calculation formula is as follows:
the Rc is calculated by the formula:
wherein n represents the number of samples, yi and yi' measured value and predicted value of the ith sample in the sample set respectively,
is the average value of the measured values of the ith sample in the sample set, wherein i is less than or equal to n.
And taking the model with the maximum correlation coefficient Rc and the minimum cross validation root mean square deviation RMSECV as an optimal model, and obtaining an optimal correction set model after comparison.
Step seven, testing the robustness of the model
In order to avoid the over-fitting phenomenon, a steady yellow smoldering time prediction model is established to achieve the purpose of practical application, therefore, all verification set samples are applied to test the effect of the artificial neural network prediction model at different yellow smoldering times, and the obtained result is represented by a correlation coefficient (Rp) and a verification mean square error (RMSEP), wherein the larger the Rp and the smaller the RMSEP, the better the model robustness is represented, and the yellow smoldering time of the sample can be accurately predicted.
Wherein the RMSEP calculation formula is as follows:
the calculation formula of Rp is as follows:
wherein n represents the number of samples, yi and yi' the measured value and the predicted value of the ith sample in the sample set are respectively, wherein i is less than or equal to n.
Specifically, in the step one, the number of the fresh leaf samples is 120, and the fresh leaf samples are randomly divided into a correction set and a verification set according to the ratio of 3:1, wherein 90 correction set samples and 30 verification set samples are obtained.
Specifically, the data of the optimum spectrum information subinterval screened and modeled in the fifth step is 6406.4-6703.4cm-1。
Specifically, in the sixth step, when Principal Component Analysis (PCA) is used to perform principal component analysis on the screened optimal spectrum information subinterval data, the first 3 principal components are used to represent the spectrum information of the optimal spectrum subinterval.
Specifically, 4 neurons and logistic information transfer functions are selected in the sixth step to establish unsupervised Kohonen structure artificial neural network models of different samples with different dark yellow time.
In conclusion, the invention provides a method for accurately predicting yellow smoldering time of a yellow tea sample by applying a near infrared spectrum technology in combination with an unsupervised Kohonen structure artificial neural network method. The model operation amount is greatly reduced, the model structure is simplified, and meanwhile, the prediction accuracy of the model and the practicability of the model are improved.
The first embodiment is as follows:
(1) fresh leaf sample collection and classification
120 fresh leaf samples of three different parts, namely single bud (without stem), one bud and one leaf (consisting of single bud, first leaf and longer stem) and one bud and two leaves (consisting of single bud, first leaf, second leaf and long stem) of Anji white tea variety in Yuan county of Hubei province are collected. And (5) after the sample is subjected to enzyme deactivation, performing smoldering yellowing, and accurately recording the smoldering yellowing time. According to different dark yellow time, the samples are divided into a correction set and a verification set by a ratio of 3:1, wherein the correction set comprises 90 samples, and the verification set comprises 30 samples, and is used for testing the robustness of a correction set model.
(2) Spectrum scanning
The near infrared spectrum of all the smoldering yellow samples is obtained by scanning with an integrating sphere diffuse reflection optical platform by using a American Saimeri-Antaris II type Fourier transform near infrared spectrometer (FT-NIR), wherein the spectrum scanning range is 4000-10000cm-1Resolution of 8cm-1And the detector is InGaAs. Spectra were acquired 3 times per sample, 64 times per scan, and the 3 acquired spectra were then averaged to give the average spectrum as the final spectrum for that sample for subsequent modeling.
Before scanning the spectrum of a sample, the near-infrared spectrometer is preheated for 30 minutes, the indoor temperature and humidity are kept basically consistent, then the sample is placed into a rotating cup matched with the instrument for spectrum scanning, the sample loading thickness of the sample is kept consistent every time, the sample cannot be penetrated by near-infrared light, and the spectrum of the whole yellow sample is shown in figure 1.
(3) Spectral noise information preprocessing
In the spectrum acquisition process, noise information which influences the model prediction effect, such as high-frequency noise, background information existing in baseline disturbance and the like, is usually generated, and therefore, the spectrum needs to be preprocessed before the correction set model is established. The near infrared spectra of all the dark yellow samples are respectively subjected to smoothing, first derivative, second derivative, multiple scattering correction and vector normalization preprocessing by using chemometrics software TQ Analyst 9.4.45 software and OPUS 7.0 software, and then each sample spectrum is converted into 1560 pairs of data points which are used for subsequent data analysis in an excel table to establish a prediction model.
By comparison, the optimal spectrum preprocessing method is vector normalization.
(4) Spectral subinterval partitioning
All spectral data points are equally divided into 20 spectral information sub-intervals, each containing 78 data points.
(5) Least squares support vector machine (LS-SVM) model building
A least squares support vector machine model of each spectrum subinterval data is established respectively, and the obtained results are shown in the following table 1:
TABLE 1 least squares support vector machine method modeling results
| Building module | Spectral interval/cm-1 | | RMSECV | |
| 1 | 3999.6-4296.6 | 0.4677 | 8.96 | |
| 2 | 4300.5-4597.5 | 0.5383 | 8.02 | |
| 3 | 4601.3-4898.3 | 0.5724 | 7.82 | |
| 4 | 4902.2-5199.1 | 0.494 | 8.31 | |
| 5 | 5203-5500 | 0.4701 | 8.78 | |
| 6 | 5503.8-5800.8 | 0.4422 | 9.01 | |
| 7 | 5804.7-6101.7 | 0.4552 | 8.94 | |
| 8 | 6105.5-6402.5 | 0.8119 | 7.58 | |
| 9 | 6406.4-6703.4 | 0.8908 | 6.55 | |
| 10 | 6707.2-7004.2 | 0.8077 | 7.94 | |
| 11 | 7008-7305 | 0.7587 | 7.01 | |
| 12 | 7308.9-7605.9 | 0.7712 | 6.95 | |
| 13 | 7609.7-7906.7 | 0.7445 | 7.11 | |
| 14 | 7910.6-8207.6 | 0.6048 | 7.53 | |
| 15 | 8211.4-8508.4 | 0.6405 | 6.93 | |
| 16 | 8512.3-8809.2 | 0.6673 | 6.63 | |
| 17 | 8813.1-9110.1 | 0.5973 | 7.94 | |
| 18 | 9113.9-9407.1 | 0.6332 | 7.04 | |
| 19 | 9410.9-9704 | 0.6531 | 7.25 | |
| 20 | 9707.9-10000 | 0.6572 | 7.05 |
It can be seen from table 1 above that, the near-infrared models of 20 sub-intervals are respectively established by applying the least square support vector machine method, and when RMSECV is minimum and the correlation coefficient Rc is maximum, the spectral interval modeled at this time is the optimal modeling sub-interval. Therefore, when 6406.4-6703.4cm-1At this time, the correlation coefficient 0.8908 and RMSECV of the model are 6.55, and the result of the least squares support vector machine model established at this time is the best, so that the best modeling spectrum subinterval is known as: 6406.4-6703.4cm-1。
(6) The method for establishing the artificial neural network prediction model of the Unstupervised Kohonen structure comprises the following steps:
1) principal component analysis of optimal spectral information subintervals
And (3) carrying out principal component analysis on the optimal spectrum subinterval data of the preprocessed yellow smoldering sample by using Matlab 2012a software to obtain the number of principal components, the contribution rate and the score value of the principal components. The contribution ratios of the first 8 principal components are shown in table 2 below:
TABLE 2 first 8 principal component contribution rates
As can be seen from table 2, the PC1 contribution rate was the largest and 90.78%, the contribution rate of the main component from PC1 to PC8 was drastically reduced, and the PC8 contribution rate was only 0.01%. The cumulative contribution rate of the three principal components PC1, PC2 and PC3 is 99.64%, and can completely represent the spectrum information of the optimal spectrum subinterval, and the first 3 principal components of the calibration set sample are scored as shown in table 3 below for the subsequent data analysis:
TABLE 3 first 3 principal component scores for the calibration set samples
2) Construction of artificial neural network prediction model of Unstupervised Kohonen structure
In order to effectively improve the robustness of the model and reduce the adverse effect of the input of noise information on the model, input variables are required to be as few as possible during modeling, but original spectral data information is required to be effectively represented, so that the artificial neural network prediction models with 9 different stuffy yellow times are established by taking the scores of the first 3 principal components analyzed and screened as input values and taking the different stuffy yellow times of the sample as output values through optimizing the number of neurons and transfer functions for many times.
When a model is established, the model prediction effect is greatly influenced due to the difference of the number of neurons in the model and the output interlayer information transfer function; therefore, when an artificial neural network model of unsupervised Kohonen structure is established, the influence of different numbers of neurons and different internal information transfer functions on the model prediction result is respectively compared, and see table 4 below. And (3) inputting the first 3 principal component scores into the artificial neural network model, and comparing the correlation coefficient Rc of the model with the interactive verification root mean square variance RMSECV value to obtain an optimal prediction model. The optimal correction set model is: slab1 has 4 neurons, the transfer function logistic, at which point the model Rc is 0.974 and the RMSECV is 2.5.
TABLE 49 artificial neural network model results
(7) Model robustness test
To prevent overfitting, the calibration set model was examined using 30 samples of the validation set and the results are expressed as the correlation coefficient Rp and the mean square error RMSEP of the validation set, see table 4 above.
As can be seen from table 4, in the artificial neural network model of unsupervised Kohonen structure of different dark yellow time samples, when the number of neurons is 3 and the transfer function is linear [0,1], the optimum correction set model Rc is 0.948 and RMSECV is 4.2, and when the robustness of the correction set model is tested by using all 30 verification set samples, the verification set model Rp is 0.934 and rmsecp is 4.8. When the number of neurons is 4 and the transfer function is logistic, the optimal calibration set model Rc is 0.974 and the RMSECV is 2.5, and when the calibration set model robustness was examined with all 30 validation set samples, the validation set model Rp was 0.965 and the RMSEP was 2.7. When the number of neurons is 5 and the transfer function is logistic, the optimum correction set model Rc is 0.965 and the RMSECV is 2.6, and when the correction set model robustness was examined with all 30 validation set samples, the validation set model Rp was 0.941 and the RMSEP was 3.9. Therefore, in the artificial neural network model established under the condition that the same unsurved Kohonen structure is applied but different neuron numbers and different information transfer functions are adopted, the prediction result of the artificial neural network model is optimal and the model prediction effect is best by using the sample artificial neural network model with different dark yellow time and established when 4 neurons and transfer functions are logistic; secondly, establishing artificial neural network models for different yellowing time samples with 5 neurons and a transfer function of logistic; the worst is the artificial neural network model with 3 neurons and different dark yellow time samples established when the transfer function is linear [0,1 ]. Therefore, in the same artificial neural network modeling method, the number of internal neurons and the information transfer function are different, and a prediction result of the model is greatly influenced, so that the number of neurons and the information transfer function need to be reasonably selected when the model is established.
The optimal artificial neural network model established using 4 neurons and transfer functions for logistic was used to predict the dark yellow time for 30 validation set samples, and the prediction results are shown in table 5 below. As can be seen from Table 5, the difference (deviation) between the true value and the predicted value of the sample yellowing time is within + -1.0, which indicates that the model predicts correctly for all samples, and the discrimination rate is 100%. Therefore, the fast and accurate prediction of the dark yellow time is realized by the unsupervised Kohonen structure artificial neural network model established when 4 neurons and transfer functions are applied as logistic.
Table 530 verification sets sample yellowing time prediction results (minutes)
In conclusion, the invention provides a method for accurately predicting yellow smoldering time of a yellow tea sample by applying a near infrared spectrum technology in combination with an unsupervised Kohonen structure artificial neural network method, firstly removing noise information of the sample to obtain an optimal spectrum preprocessing method, namely vector normalization, and converting a sample spectrum into paired data points to be stored in an excel table; then all the spectral data are divided into 20 subintervals, and least square branches of each subinterval data are respectively establishedA support vector machine method model, screening out the best subinterval data (6406.4-6703.4 cm)-15.00% of the total spectral data); screening the first 3 principal components by applying principal component analysis, wherein the accumulated contribution rate of the first 3 principal components is 99.64%, establishing an unsupervised Kohonen structure artificial neural network prediction model with different neuron numbers and information transfer functions by taking the scores of the first 3 principal components as input values, and establishing an unsupervised Kohonen structure artificial neural network model with 4 neurons and transfer functions as logistic with the best prediction result (Rp is 0.965, SERMP is 2.7); secondly, an unsupervised Kohonen structure artificial neural network model which is established when 5 neurons and a transfer function are logistic; worst with 3 neurons and a transfer function of linear [0,1]]An unsupervised Kohonen structure artificial neural network model is built. Therefore, the method combines a least square support vector machine method and an unsupervised Kohonen structure artificial neural network method, perfectly realizes the accurate prediction of the yellow smoldering time of the Yuan' an yellow tea sample processed by three picking standards of single bud, one bud and one leaf and one bud and two leaves (the prediction deviation is all in the range of +/-1.0, the prediction accuracy is 100%), and the established prediction model not only achieves the purposes of greatly reducing the model calculation amount (the modeling spectrum data accounts for 5.00% of all the spectrum data) and simplifying the model, but also achieves the purposes of improving the prediction accuracy of the model and enhancing the practicability of the model.
The foregoing is a further detailed description of the invention in connection with specific preferred embodiments and it is not intended to limit the invention to the specific embodiments described. For those skilled in the art to which the invention relates, several simple deductions or substitutions may be made without departing from the spirit of the invention, and the above-mentioned structures should be considered as belonging to the protection scope of the invention.
Claims (2)
1. A method for rapidly predicting the time of smoldering yellow of far-safe yellow tea by applying a near infrared spectrum technology is characterized in that the near infrared spectrum of fresh leaf samples at different smoldering yellow times is obtained by scanning, and after the spectrum of the sample is preprocessed to remove noise information, the spectrum of the sample is converted into paired data points; all the spectral data are divided into 20 subintervals, a least square support vector machine method model of each subinterval data is respectively established, and the best subinterval data for modeling are screened out; and then compressing and extracting the optimal subinterval information data points by using a principal component analysis method, establishing artificial neural network prediction models of different yellow-smoldering time samples by using the principal component score as an input numerical value and regulating the number of neurons in a hidden layer and an information transfer function, wherein the artificial neural network prediction models are used for predicting the yellow-smoldering time of the yellow tea, and the method specifically comprises the following steps:
step one, fresh leaf sample collection and classification
Collecting three fresh leaf samples of single bud, one bud and one leaf and one bud and two leaves at different positions of an Anji white tea variety in Yuan-an county of Hubei province, deactivating enzymes, performing smoldering yellowing on the samples, and simultaneously accurately recording smoldering yellowing time; according to different dark yellow time, dividing the sample into 2 sets of a correction set and a verification set, and respectively establishing a correction set near infrared spectrum prediction model and testing the robustness of the correction set prediction model;
wherein, one bud and one leaf are composed of single bud, first leaf and long stalk, and one bud and two leaves are composed of single bud, first leaf, second leaf and long stalk;
step two, spectrum scanning
Scanning by a near-infrared spectrometer to obtain the near-infrared spectrum of all the smoldering yellow samples, wherein the spectrum scanning range is 4000-10000cm-1Resolution of 8cm-1The detector is InGaAs, 3 spectra are collected by each sample, each time of scanning is carried out for 64 times, then the 3 collected spectra are averaged, and the average spectrum is used as the final spectrum of the sample for subsequent modeling;
thirdly, preprocessing the spectral noise information
Performing denoising pretreatment on the near infrared spectrum obtained in the step two by using chemometrics software by using a vector normalization method, so that the signal-to-noise ratio of the spectrum is improved, and a stable prediction model is favorably established; after the spectrum is denoised, converting the sample spectrum into paired data points;
step four, spectrum subinterval division
Dividing all spectral information data into 20 spectral information subintervals equally, and accurately screening the spectral information subintervals reflecting the sample yellowing time for subsequently establishing a least square support vector machine model;
step five, establishing an LS-SVM model of a least square support vector machine
The invention respectively establishes a prediction model of each spectral information subinterval by applying an LS-SVM method, compares a model correlation coefficient Rc with the magnitude of a cross validation root mean square variance RMSECV, preliminarily screens out the best spectral information subinterval data of the modeling, and achieves the purpose of accurately screening the spectral information reflecting the yellow-smoldering time, wherein the best spectral information subinterval data of the modeling screened out in the step is 6406.4-6703.4cm-1(ii) a Wherein Rc is the largest, RMSECV is the smallest, which means that the result of the established least square support vector machine model is the best,
wherein, RMSECV formula is:
the Rc is calculated by the formula:
wherein n represents the number of samples, yi and yi' measured value and predicted value of the ith sample in the sample set respectively,
is the average value of the measured values of the ith sample in the sample set, wherein i is less than or equal to n;
step six, establishing unsupervised Kohonen structure artificial neural network prediction model
Further accurately predicting the yellowing time of the sample by using a nonlinear artificial neural network method, wherein the method comprises the following steps:
1) principal component analysis of optimal spectral information subintervals
Carrying out principal component analysis on the screened optimal spectrum information subinterval data by adopting a Principal Component Analysis (PCA) method to obtain an individual contribution rate value, an accumulated contribution rate value and a principal component score of each principal component; the sample modeling spectrum subinterval information can be effectively represented only when the principal component accumulated contribution rate is more than or equal to 85%; in the step, when principal component analysis is carried out on the screened optimal spectrum information subinterval data by adopting a Principal Component Analysis (PCA), the first 3 principal components are adopted to represent the spectrum information of the optimal spectrum subinterval;
2) artificial neural network prediction model establishment
Establishing an artificial neural network prediction model of an unsupervised Kohonen structure by using Neuro Shell 2 software and taking the principal component score of an optimal spectrum information subinterval as an input value and different yellowing times of a sample as an output value, wherein the unsupervised Kohonen structure artificial neural network comprises three information transfer functions of 1 hidden layer, 3, 4 and 5 neuron numbers, logistic, linear [0,1] and tanh, and the magnitude of a model correlation coefficient Rc and a cross validation root mean square deviation RMSECV are compared to obtain the optimal near infrared spectrum prediction model, wherein the larger the Rc and the smaller the RMSECV are, the better the prediction effect of the model is represented; selecting 4 neurons and a logistic information transfer function to establish unsupervised Kohonen structure artificial neural network models of different samples at different smoldering times;
wherein the RMSECV calculation formula is as follows:
the Rc is calculated by the formula:
wherein n represents the number of samples, yi and yi' measured value and predicted value of the ith sample in the sample set respectively,
the average value of the measured values of the ith sample in the sample set is shown, wherein i is less than or equal to n;
the model with the largest correlation coefficient Rc and the smallest cross validation root-mean-square variance RMSECV is used as an optimal model, and an optimal correction set model is obtained after comparison;
step seven, testing the robustness of the model
The artificial neural network prediction model effects of different yellowing times are tested by applying all verification set samples, and the obtained result is represented by a correlation coefficient Rp and a verification mean square error RMSEP, wherein the larger the Rp is, the smaller the RMSEP is, the better the model robustness is, and the yellowing time of the sample can be accurately predicted;
wherein the RMSEP calculation formula is as follows:
the calculation formula of Rp is as follows:
wherein n represents the number of samples, yi and yi' the measured value and the predicted value of the ith sample in the sample set are respectively, wherein i is less than or equal to n.
2. The method for rapidly predicting the time for smoldering yellow tea by using the near infrared spectrum technology as claimed in claim 1, wherein the method comprises the following steps: in the first step, the number of the fresh leaf samples is 120, the fresh leaf samples are divided into a correction set and a verification set according to the ratio of 3:1, wherein 90 correction set samples and 30 verification set samples are obtained.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201910513186.3A CN110308111B (en) | 2019-06-14 | 2019-06-14 | Method for rapidly predicting time for smoldering yellow tea by using near infrared spectrum technology |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201910513186.3A CN110308111B (en) | 2019-06-14 | 2019-06-14 | Method for rapidly predicting time for smoldering yellow tea by using near infrared spectrum technology |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN110308111A CN110308111A (en) | 2019-10-08 |
| CN110308111B true CN110308111B (en) | 2022-05-06 |
Family
ID=68077228
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201910513186.3A Active CN110308111B (en) | 2019-06-14 | 2019-06-14 | Method for rapidly predicting time for smoldering yellow tea by using near infrared spectrum technology |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN110308111B (en) |
Families Citing this family (3)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN112861412A (en) * | 2019-11-27 | 2021-05-28 | 国能生物发电集团有限公司 | Biomass volatile component content measurement and modeling method based on near infrared spectrum principal component and neural network |
| CN114624402B (en) * | 2022-01-28 | 2023-06-27 | 广西壮族自治区水产科学研究院 | Quality evaluation method for snail rice noodle sour bamboo shoots based on near infrared spectrum |
| CN116793991B (en) * | 2023-08-22 | 2023-11-10 | 青岛理工大学 | Glutamic acid concentration measurement method based on near infrared spectrum and mixing loss |
Family Cites Families (19)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| US5175795A (en) * | 1988-07-29 | 1992-12-29 | Hitachi, Ltd. | Hybridized frame inference and fuzzy reasoning system and method |
| CN101059425A (en) * | 2007-05-29 | 2007-10-24 | 浙江大学 | Method and device for identifying different variety green tea based on multiple spectrum image texture analysis |
| US8158175B2 (en) * | 2008-08-28 | 2012-04-17 | Frito-Lay North America, Inc. | Method for real time measurement of acrylamide in a food product |
| CN102507457A (en) * | 2011-11-18 | 2012-06-20 | 江苏大学 | Device and method for rapidly and nondestructively detecting crop nutrient elements |
| CN103743697A (en) * | 2013-12-20 | 2014-04-23 | 贵州省分析测试研究院 | Method for monitoring tea production in real time by adopting near infrared spectrum |
| CN104951990B (en) * | 2014-03-28 | 2018-06-05 | 国际商业机器公司 | Power grid data processing method and equipment |
| CN104020129A (en) * | 2014-05-16 | 2014-09-03 | 安徽农业大学 | Method for discriminating fermentation quality of congou black tea based on near-infrared-spectroscopy-combined amino acid analysis technology |
| US10641758B2 (en) * | 2015-09-01 | 2020-05-05 | Exxonmobil Upstream Research Company | Apparatus, systems, and methods for enhancing hydrocarbon extraction and techniques related thereto |
| 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 |
| CN106290240A (en) * | 2016-08-29 | 2017-01-04 | 江苏大学 | A kind of method based on near-infrared spectral analysis technology to Yeast Growth curve determination |
| CN106560700A (en) * | 2016-10-20 | 2017-04-12 | 中国计量大学 | Machine learning method for identifying origin of Wuyi rock tea automatically |
| CN106525849B (en) * | 2016-11-02 | 2019-03-05 | 江苏大学 | The intelligent assorted method and system of tealeaves |
| CN206651317U (en) * | 2017-02-28 | 2017-11-21 | 四川俊龙农业科技有限公司 | A kind of tealeaves high efficiency smart steaming apparatus |
| CN107348021B (en) * | 2017-06-14 | 2020-05-19 | 中国农业科学院茶叶研究所 | Yellow tea smoldering degree control method based on color difference system |
| CN107958267B (en) * | 2017-11-21 | 2021-04-27 | 东南大学 | Oil product property prediction method based on spectral linear representation |
| CN107860740A (en) * | 2017-12-08 | 2018-03-30 | 中国农业科学院茶叶研究所 | A kind of evaluation method of the fermentation of black tea quality based on near-infrared spectrum technique |
| CN108872132A (en) * | 2018-08-24 | 2018-11-23 | 湖北省农业科学院果树茶叶研究所 | A method of fresh tea leaves kind is differentiated using near infrared spectrum |
| CN109001147A (en) * | 2018-08-24 | 2018-12-14 | 湖北省农业科学院果树茶叶研究所 | A method of fresh tea leaves geography information is differentiated using near infrared spectrum |
| CN109685098B (en) * | 2018-11-12 | 2024-03-19 | 江苏大学 | Tea variety classification method for fuzzy inter-cluster separation and clustering |
-
2019
- 2019-06-14 CN CN201910513186.3A patent/CN110308111B/en active Active
Also Published As
| Publication number | Publication date |
|---|---|
| CN110308111A (en) | 2019-10-08 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| Chandrasekaran et al. | Potential of near-infrared (NIR) spectroscopy and hyperspectral imaging for quality and safety assessment of fruits: An overview | |
| dos Santos et al. | A review on the application of vibrational spectroscopy in the wine industry: From soil to bottle | |
| CN110308111B (en) | Method for rapidly predicting time for smoldering yellow tea by using near infrared spectrum technology | |
| JP6339244B2 (en) | Method for predicting sugar content and acidity of fruit using multivariate statistical analysis of FT-IR spectrum data | |
| Gao et al. | Hyperspectral image information fusion-based detection of soluble solids content in red globe grapes | |
| Abdullah et al. | Automated inspection system for colour and shape grading of starfruit (Averrhoa carambola L.) using machine vision sensor | |
| CN106568738A (en) | Method of using near infrared spectroscopy to rapidly determine fresh leaves of tea in different quality grades | |
| CN109001147A (en) | A method of fresh tea leaves geography information is differentiated using near infrared spectrum | |
| Ioannou‐Papayianni et al. | Authenticity of Cypriot sweet wine commandaria using FT‐IR and chemometrics | |
| CN112669915A (en) | Pear nondestructive testing method based on neural network and near infrared spectrum | |
| CN110987865A (en) | Method for detecting fig quality based on near infrared spectrum | |
| CN113033066A (en) | Method for establishing near infrared spectrum identification model of sargassum fusiforme production area, strain and cultivation mode and identification method | |
| CN111795943A (en) | Method for nondestructive detection of exogenous doped sucrose in tea based on near infrared spectrum technology | |
| CN114092839B (en) | Unmanned aerial vehicle remote sensing-based soybean harvest period maturity judging method | |
| CN110320174B (en) | Method for rapidly predicting time for smoldering yellow tea by applying polynomial net structure artificial neural network | |
| Luo et al. | Quantitative detection of soluble solids content, pH, and total phenol in Cabernet Sauvignon grapes based on near infrared spectroscopy | |
| CN113751345B (en) | Pineapple nondestructive testing and grading device and method | |
| Yang | Nondestructive prediction of optimal harvest time of cherry tomatoes using VIS-NIR spectroscopy and PLSR calibration | |
| Li et al. | Two-wavelength image detection of early decayed oranges by coupling spectral classification with image processing | |
| CN110308110B (en) | Nondestructive prediction method for long-distance safe yellow tea yellow-smoldering time based on least square support vector machine | |
| CN110186870B (en) | Method for distinguishing fresh tea leaf producing area of Enshi Yulu tea by extreme learning machine spectrum model | |
| CN107796779A (en) | The near infrared spectrum diagnostic method of rubber tree LTN content | |
| CN116678855A (en) | Method for evaluating freshness of sunflower seeds by utilizing hyperspectrum of black skin of sunflower seeds | |
| Fu et al. | Discrimination of pear varieties using three classification methods based on near-infrared spectroscopy | |
| CN110334714A (en) | A kind of machine based on artificial neural network technology adopts mee tea vehicle tinctorial pattern product grade prediction technique |
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 | ||
| GR01 | Patent grant | ||
| GR01 | Patent grant |