CN106568738A - Method of using near infrared spectroscopy to rapidly determine fresh leaves of tea in different quality grades - Google Patents
Method of using near infrared spectroscopy to rapidly determine fresh leaves of tea in different quality grades Download PDFInfo
- Publication number
- CN106568738A CN106568738A CN201610930404.XA CN201610930404A CN106568738A CN 106568738 A CN106568738 A CN 106568738A CN 201610930404 A CN201610930404 A CN 201610930404A CN 106568738 A CN106568738 A CN 106568738A
- Authority
- CN
- China
- Prior art keywords
- sample
- fresh leaf
- near infrared
- different quality
- fresh
- 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
Links
- 241001122767 Theaceae Species 0.000 title claims abstract description 42
- 238000000034 method Methods 0.000 title claims abstract description 31
- 238000004497 NIR spectroscopy Methods 0.000 title abstract 3
- 238000001228 spectrum Methods 0.000 claims abstract description 47
- 238000002329 infrared spectrum Methods 0.000 claims abstract description 28
- 238000013528 artificial neural network Methods 0.000 claims abstract description 19
- 238000000513 principal component analysis Methods 0.000 claims description 15
- 239000000470 constituent Substances 0.000 claims description 11
- 230000000694 effects Effects 0.000 claims description 11
- 238000012546 transfer Methods 0.000 claims description 8
- 230000003595 spectral effect Effects 0.000 claims description 7
- 238000004611 spectroscopical analysis Methods 0.000 claims description 5
- 238000012360 testing method Methods 0.000 claims description 5
- 229910000530 Gallium indium arsenide Inorganic materials 0.000 claims description 4
- 238000009795 derivation Methods 0.000 claims description 4
- 238000012795 verification Methods 0.000 claims description 4
- 238000002512 chemotherapy Methods 0.000 claims description 3
- 238000005259 measurement Methods 0.000 claims 1
- 230000005540 biological transmission Effects 0.000 abstract description 6
- 238000007781 pre-processing Methods 0.000 abstract 1
- 210000005036 nerve Anatomy 0.000 description 7
- 238000012545 processing Methods 0.000 description 4
- 241000607479 Yersinia pestis Species 0.000 description 2
- RYYVLZVUVIJVGH-UHFFFAOYSA-N caffeine Chemical compound CN1C(=O)N(C)C(=O)C2=C1N=CN2C RYYVLZVUVIJVGH-UHFFFAOYSA-N 0.000 description 2
- 238000007405 data analysis Methods 0.000 description 2
- 230000004069 differentiation Effects 0.000 description 2
- 201000010099 disease Diseases 0.000 description 2
- 208000037265 diseases, disorders, signs and symptoms Diseases 0.000 description 2
- 230000003993 interaction Effects 0.000 description 2
- 210000002569 neuron Anatomy 0.000 description 2
- 239000002994 raw material Substances 0.000 description 2
- 210000000697 sensory organ Anatomy 0.000 description 2
- 241001269238 Data Species 0.000 description 1
- LPHGQDQBBGAPDZ-UHFFFAOYSA-N Isocaffeine Natural products CN1C(=O)N(C)C(=O)C2=C1N(C)C=N2 LPHGQDQBBGAPDZ-UHFFFAOYSA-N 0.000 description 1
- 244000061176 Nicotiana tabacum Species 0.000 description 1
- 235000002637 Nicotiana tabacum Nutrition 0.000 description 1
- 230000009286 beneficial effect Effects 0.000 description 1
- 229960001948 caffeine Drugs 0.000 description 1
- VJEONQKOZGKCAK-UHFFFAOYSA-N caffeine Natural products CN1C(=O)N(C)C(=O)C2=C1C=CN2C VJEONQKOZGKCAK-UHFFFAOYSA-N 0.000 description 1
- 238000012937 correction Methods 0.000 description 1
- 238000010586 diagram Methods 0.000 description 1
- 239000003814 drug Substances 0.000 description 1
- 235000013399 edible fruits Nutrition 0.000 description 1
- 238000005516 engineering process Methods 0.000 description 1
- 238000003306 harvesting Methods 0.000 description 1
- 238000007689 inspection Methods 0.000 description 1
- 238000003062 neural network model Methods 0.000 description 1
- 230000007935 neutral effect Effects 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
- 238000012216 screening Methods 0.000 description 1
- 230000035945 sensitivity Effects 0.000 description 1
- 239000004753 textile Substances 0.000 description 1
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
- 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
- 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
- Physics & Mathematics (AREA)
- Spectroscopy & Molecular Physics (AREA)
- Health & Medical Sciences (AREA)
- Life Sciences & Earth Sciences (AREA)
- Chemical & Material Sciences (AREA)
- Analytical Chemistry (AREA)
- Biochemistry (AREA)
- General Health & Medical Sciences (AREA)
- General Physics & Mathematics (AREA)
- Immunology (AREA)
- Pathology (AREA)
- Investigating Or Analysing Materials By Optical Means (AREA)
Abstract
The invention discloses a method of using near infrared spectroscopy to rapidly determine fresh leaves of tea in different quality grades. The method comprises the following steps: using near infrared spectroscopy scanning to obtain the near infrared spectrums of samples of fresh tea leaves in different quality grades; analyzing the main components of the fresh leaves according to the spectrums; and inputting the values of the main components to establish an artificial neural network predictive model with multiple information transmission modes. Specifically, the method comprises the following steps: collecting and classifying fresh leaf samples, collecting spectrums, preprocessing the spectrums, analyzing the main components of fresh leaves, establishing an artificial neural network predictive model, and verifying the model. The method has the advantages that the physical states of fresh leaves are maintained, the fresh leaves of tea in different quality grades are determined rapidly, accurately and non-destructively, and the method also guarantees the stable quality of tea.
Description
Technical field
The present invention relates to a kind of decision method of different quality grade tea fresh leaves, more specifically to a kind of near infrared light
The quick method for judging different quality grade tea fresh leaves of spectrum.
Background technology
, in processing, the requirement to leaf quality is very strict for YULU of bestowing favour tea.Superfine YULU tea of bestowing favour is divided into superfine one
Deng and superfine second-class two ranks, the YULU tea requirement fresh leaf wherein superfine first-class is bestowed favour:Simple bud is more than 95% or lobule kind one
Open up at the beginning of one leaf of bud, raw material is fresh, even neat, without red change bud-leaf, without Purple tea shoots, disease pest bud-leaf, rainwater bud-leaf;Superfine is second-class to bestow favour
YULU tea requires fresh leaf:Open up at the beginning of one bud, one leaf more than 95%, raw material is fresh, neat, without red stalk bud-leaf, Purple tea shoots, disease pest bud
Leaf, rainwater bud-leaf.
Superfine bestows favour YULU tea before processing, Tea Processing personnel typically by itself working experience and sense organ method come
Judge leaf quality grade, but the sensory organ sensitivity of people easily by external worlds such as physiological situation at that time and weather, humitures
The impact of condition, with larger subjectivity and randomness, the YULU tea leaf quality that is unfavorable for bestowing favour is stablized.Therefore, it is badly in need of building
A kind of method of vertical quick, accurate, lossless judgement processing leaf quality grade.
The content of the invention
Present invention aims to existing tea quality grade judges to lack with larger subjectivity and randomness etc.
Fall into, there is provided a kind of method that near infrared spectrum quickly judges different quality grade tea fresh leaves.
For achieving the above object, technical solution of the invention is:A kind of near infrared spectrum quickly judges different quality
The method of grade tea fresh leaves, scanning obtain the fresh leaf sample near infrared spectrum of different quality grade, then to fresh leaf sample spectra
Carry out principal component analysiss, then with main constituent as input value set up much information transfer mode different quality grade fresh leaf it is artificial
Neural network prediction model judges the different quality grade of fresh leaf, specifically includes following steps:
Step one, fresh leaf sample collecting and classification
Two kinds of different quality grades tea fresh leaves sample of the same race is gathered respectively, according to the difference of credit rating, by tea fresh leaves sample
Product random division is 2 set of calibration set and checking collection;
Step 2, spectra collection
The near infrared spectrum for obtaining whole fresh leaf samples is scanned using Fourier-type near infrared spectrometer;
Step 3, Pretreated spectra
Applied Chemometrics software carries out derivation and smooth pretreatment to the near infrared spectrum of whole fresh leaf samples, then
Fresh leaf sample spectra is converted into into paired data point;
Step 4, fresh leaf sample spectra principal component analysiss
Principal component analysiss are carried out to the spectroscopic data of whole fresh leaf samples using Matlab softwares, whole fresh leaf samples are tried to achieve
The score Score1 value and Score2 values, number of principal components and its contribution rate of spectroscopic data;
Step 5, set up Artificial Neural Network Prediction Model
Front 3 main constituents with calibration set sample spectra as input value, with different quality grade fresh leaf type as output valve,
Through optimizing repeatedly, standard nets, tri- kinds of letters of jump connection nets and Jordan-Elman nets are set up
Breath transfer mode fresh leaf different quality grade Artificial Neural Network Prediction Model, compares three kinds of model coefficient R and interaction
Checking root-mean-square variance RMSECV value,
Wherein coefficient R formula is:
Validation-cross root-mean-square variance RMSECV formula is:
In formula, R is correlation coefficient, and n represents sample number, yiAnd yi' it is respectively the credit rating of i-th sample in sample sets
Measured value and credit rating predictive value,For the meansigma methodss of the measured value of i-th sample in sample sets, i≤n in formula,
Wherein with the maximum model minimum with validation-cross root-mean-square variance RMSECV of coefficient R as best model, Jing
After obtain optimal calibration set model;
Step 6, model checking
To avoid the occurrence of overfitting phenomenon, application verification collection sample enters to the three kinds of calibration set forecast result of model for obtaining
Performing check, acquired results coefficient R and checking collection mean square deviation RMSEP represent that wherein coefficient R is bigger and checking collection is equal
Variance RMSEP is more little then to represent that test effect is better, if the credit rating predictive value of the near infrared spectrum for now obtaining and quality
Grade measured value is basically identical, then it represents that fine to the prediction effect of checking collection sample, and optimal calibration set model can be accurate
The different quality grade of prediction fresh leaf sample,
Wherein checking collection mean square deviation RMSEP formula is:
In formula, n represents sample number, yiAnd yi' it is respectively the credit rating measured value and quality of i-th sample in sample sets
Grade forecast value, i≤n in formula.
In described step one, fresh leaf sample size is 100 parts, and wherein tea fresh leaves sample is superfine first-class and superfine second-class each
50, fresh leaf sample is according to 7:3 ratio random division is calibration set and checking collection.
The fresh leaf sample plucked in described step one is to open up at the beginning of bud, one leaf of a bud.
Fourier-type near infrared spectrometer in the step 2 is with silent winged your II type Fu of Antaris of generation of U.S.'s match
Leaf near infrared spectrometer, spectral scanning range 4000-10000cm-1, resolution 8cm-1, detector is InGaAs, each sample
10 spectrum of collection, every time scanning 64 times take the final spectrum of the meansigma methodss as the sample of 10 collection spectrum.
Chemo metric software in the step 3 is 7.0 software of TQ Analyst 9.4.45 softwares and OPUS.
Compared with prior art, beneficial effects of the present invention:
Near-infrared spectrum technique is based in the present invention, with reference to principal component analysiss and the artificial neuron of much information transfer mode
Network model judges leaf quality grade, realizes different quality grade tea fresh under conditions of intact guarantee fresh leaf physical state
Quick, accurate, the lossless judgement of leaf, the fresh leaf for effectively solving its different quality grade when fresh leaf is purchased judge difficult asking
Topic, result of study also provide a kind of safeguard of science for the stay in grade of Folium Camelliae sinensis.
Description of the drawings
Fig. 1 is whole 100 fresh leaf sample spectrum diagrams in the present invention.
Fig. 2 is superfine first-class fresh leaf and superfine second-class fresh leaf sample Scores1 values and Scores2 values space point in the present invention
Butut.
Fig. 3 is standard nets information transmission artificial neural network structures in the present invention.
Fig. 4 is jump connection nets information transmission artificial neural network structures in the present invention.
Fig. 5 is Jordan-Elman nets information transmission artificial neural network structures in the present invention.
Specific embodiment
Below in conjunction with description of the drawings, the present invention is described in further detail with specific embodiment.
A kind of method that near infrared spectrum quickly judges different quality grade tea fresh leaves, scanning obtain different quality grade
Fresh leaf sample near infrared spectrum, then carries out principal component analysiss, then sets up many as input value with main constituent to fresh leaf sample spectra
The Artificial Neural Network Prediction Model for planting the different quality grade fresh leaf of the mode of intelligence transmission judges the different quality grade of fresh leaf.
Specifically include following steps:
Step one, fresh leaf sample collecting and classification
Two kinds of different quality grades tea fresh leaves sample of the same race is gathered respectively, according to the difference of credit rating, by tea fresh leaves sample
Product random division is 2 set of calibration set and checking collection;Wherein checking collection fresh leaf sample is used to check fresh leaf different quality grade
The robustness of calibration set forecast model.
Step 2, spectra collection
The near infrared spectrum for obtaining whole fresh leaf samples is scanned using Fourier-type near infrared spectrometer (FT-NIR).
Near infrared spectrum (NIRS) is a kind of electromagnetic wave between visible region and mid-infrared light area, with it is quick,
Accurately and without the need for, the features such as pretreatment, having been widely used for agricultural, petrochemical industry, textile industry, pharmaceuticals industry and Nicotiana tabacum L. at present
In industry.In Folium Camelliae sinensis application, near-infrared spectrum technique successfully realized to caffeine, tea polyphenols total amount it is pre-
Survey and the judgement on ground that tea is traced to the source etc..
Step 3, Pretreated spectra
Applied Chemometrics software carries out derivation and smooths waiting pretreatment to the near infrared spectrum of whole fresh leaf samples, so
Fresh leaf sample spectra is converted into into paired data point afterwards, for subsequently setting up fresh leaf different quality level correction collection forecast model
With checking collection model.
Step 4, fresh leaf sample spectra principal component analysiss (PCA)
Principal component analysiss are carried out to the spectroscopic data of whole fresh leaf samples using Matlab softwares, whole fresh leaf samples are tried to achieve
The score Score1 value and Score2 values, number of principal components and its contribution rate of spectroscopic data.
Step 5, set up artificial neural network (BP-ANN) forecast model
Front 3 main constituents with calibration set sample spectra as input value, with different quality grade fresh leaf type as output valve,
Through optimizing repeatedly, standard nets, tri- kinds of letters of jump connection nets and Jordan-Elman nets are set up
Breath transfer mode fresh leaf different quality grade Artificial Neural Network Prediction Model, compares three kinds of model coefficient R and interaction
Checking root-mean-square variance RMSECV value,
Wherein coefficient R formula is:
Validation-cross root-mean-square variance RMSECV formula is:
In formula, R is correlation coefficient, and n represents sample number, yiAnd yi' it is respectively the credit rating of i-th sample in sample sets
Measured value and credit rating predictive value,For the meansigma methodss of the measured value of i-th sample in sample sets, i≤n in formula;
Wherein the model of and validation-cross root-mean-square variance RMSECV minimum maximum with coefficient R, should as best model
Model accuracy highest, obtains optimal calibration set model Jing after relatively.
Step 6, model checking
To avoid the occurrence of overfitting phenomenon, application verification collection sample enters to the three kinds of calibration set forecast result of model for obtaining
Performing check, is to verify the leaf quality grade forecast value that collect sample to predict with the three kinds of calibration set forecast models for having obtained
It is whether whether consistent with the measured value having already known.Acquired results coefficient R and checking collect mean square deviation RMSEP and represent, its
Middle coefficient R is bigger and the more little then expression test effect of checking collection mean square deviation RMSEP is better;As a result with the number of checking collection sample
According to being expressed, if the credit rating predictive value of the near infrared spectrum for now obtaining is basically identical with credit rating measured value,
Represent the prediction effect to checking collection sample very well, optimal calibration set model can accurately predict the different quality of fresh leaf sample
Grade.
Wherein checking collection mean square deviation RMSEP formula is:
In formula, n represents sample number, yiAnd yi' it is respectively the credit rating measured value and quality of i-th sample in sample sets
Grade forecast value, i≤n in formula.
Specifically, in described step one, tea fresh leaves sample size is 100 parts, and wherein tea fresh leaves sample is superfine first-class and special
Level is second-class each 50, and fresh leaf sample is according to 7:3 ratio random division is calibration set and checking collection.
Specifically, the fresh leaf sample plucked in described step one is to open up at the beginning of bud, one leaf of a bud.
Specifically, the Fourier-type near infrared spectrometer in the step 2 is with the silent winged generation that Antaris of U.S.'s match
II type Fourier transform near infrared instrument, spectral scanning range 4000-10000cm-1, resolution 8cm-1, detector is InGaAs,
10 spectrum of each sample collecting, every time scanning 64 times take the final spectrum of the meansigma methodss as the sample of 10 collection spectrum.
Specifically, the triple chemo metric software of the step is TQ Analyst 9.4.45 softwares and OPUS 7.0
Software.
Specific embodiment one:
(1) fresh leaf sample collecting and classification
Totally 100, two kinds of different quality grades tea fresh leaves sample of the same race is gathered respectively, wherein YULU of bestowing favour is superfine first-class fresh
Leaf and superfine second-class fresh leaf are each 50.Plucking time is 27 days-April 3 March in 2015;The fresh leaf sample of harvesting be bud, a bud
Open up at the beginning of one leaf.According to the difference of credit rating, it is 2 set of calibration set and checking collection by tea fresh leaves sample random division, wherein
70 samples of calibration set (superfine first-class fresh leaf and superfine second-class fresh leaf are each 35);Checking 30, sample of collection (superfine first-class fresh leaf
It is each 15 with superfine second-class fresh leaf), checking collection is used for the robustness for checking calibration set model.
(2) spectra collection
Referring to Fig. 1, using silent winged generation that II types Fourier transform near infrared instrument (FT-NIR) of Antaris of U.S.'s match, choosing
With integrating sphere diffuse-reflectance optical table;Spectral scanning range 4000-10000cm-1;Resolution 8cm-1, detector is InGaAs.
10 spectrum of each sample collecting, every time scanning 64 times take the final spectrum of the meansigma methodss as the sample of 10 collection spectrum.
Before spectra collection, the spectrogrph is preheated into 1h, after keeping indoor temperature and humidity basically identical, fresh leaf sample is loaded and the instrument
Spectrum is gathered in the supporting rotating cup of device, whole fresh leaf sample spectras are referring to Fig. 1.
(3) Pretreated spectra
During spectra collection, it will usually which producing high-frequency noise and baseline drift etc. affects the noise of forecast result of model
Information, therefore, need to carry out pretreatment to spectrum before calibration set model is set up.Therefore Applied Chemometrics software TQ
Analyst9.4.45 softwares and 7.0 softwares of OPUS carry out derivation and smooth wait pre- place to the near infrared spectrum of whole fresh leaf samples
Reason;Then fresh leaf sample spectra is converted into into 1557 pairs of data points, for subsequent data analysis, sets up discrimination model.
(4) fresh leaf spectrum principal component analysiss (PCA)
Principal component analysiss are carried out to whole fresh leaf spectrum using Matlab softwares, number of principal components and its contribution rate is tried to achieve.Front 8
The contribution rate difference of individual main constituent is as follows:
1 front 8 principal component contributor rate of table
As it can be seen from table 1 PC1 contribution rates are maximum, it is 92.40%, drastically reduces from PC1-PC8 principal component contributor rates,
PC6-PC8 contribution rates are only 0.01%.Wherein, the contribution rate of accumulative total of tri- main constituents of PC1, PC2 and PC3 is 99.87%, completely
Above-mentioned spectral information can be represented, for subsequent data analysis.
The score Score1 value and Score2 values of the whole fresh leaf sample spectral datas tried to achieve according to above-mentioned principal component analysiss
Information, obtains the locus scattergram of superfine first-class fresh leaf and superfine second-class fresh leaf sample, referring specifically to Fig. 2.
Figure it is seen that superfine first-class fresh leaf and superfine second-class fresh leaf sample are more intensive to be distributed in coordinate axess
Four quadrants, the different grades of fresh leaf sample of two classes cannot almost separate, complete weave in, no obvious boundary.
Therefore, sample scores1 values and scores2 values are tried to achieve using principal component analysiss only and then determines that the method for its locus cannot
Reach the purpose for accurately differentiating leaf quality grade.
(5) set up artificial neural network (BP-ANN) forecast model
When artificial nerve network model is set up, it is desirable to reduce input variable as far as possible, but also want generation as much as possible
Table original spectral data information, therefore, (contribution rate of accumulative total is to select front 3 main constituents for screening with above-mentioned PCA
99.87%) it is input value, (the superfine first-class fresh leaf sample with superfine first-class fresh leaf and superfine second-class fresh leaf sample type as output valve
Performance number is 1.0000, and superfine second-class fresh leaf sample value is 2.0000), through optimizing repeatedly, to set up different quality grade fresh leaf
Artificial Neural Network Prediction Model.During model is set up, due to artificial nerve network model internal information transfer mode
Difference, and cause the prediction effect for setting up model also produce larger difference.In modeling process, it is respectively compared
Standard nets, tri- kinds of mode of intelligence transmission artificial neurons of jump connection nets and Jordan-Elman nets
The prediction effect of network model, referring specifically to Fig. 3, by front 3 main constituents are separately input to 3 kinds of artificial nerve network models
In, compare three kinds of model coefficient R and validation-cross root-mean-square variance RMSECV value, obtain optimal calibration set prediction mould
Type.Optimal calibration set model is Jordan-Elman nets transfer mode artificial nerve network models, and R is 0.962, RMSECV
For 0.206.
(6) model checking
To avoid the occurrence of overfitting phenomenon, 30 parts of samples of application verification collection are tested to three kinds of calibration set models, institute
Obtain result coefficient R and checking collection mean square deviation RMSEP is represented, referring specifically to table 2 below:
23 kinds of artificial nerve network model modeling results of table compare
From table 2 it can be seen that superfine first-class and superfine second-class fresh leaf sample standard nets structure artificial neutral nets
Model calibration set coefficient R is 0.665, and validation-cross root-mean-square variance RMSECV is 0.403, is carried out when sample is collected with checking
During inspection, it is 0.435 for 0.612, RMSEP to be verified collection model R.It is superfine first-class and superfine second-class fresh superfine first-class and superfine
Second-class fresh leaf jump connection nets structure artificials neural network model calibration set R is 0.395 for 0.687, RMSECV,
When being tested with checking collection sample, it is 0.41 for 0.635, RMSEP to be verified collection model R.Superfine first-class and superfine two
Be 0.206 for 0.962, RMSECV etc. fresh leaf Jordan-Elman nets structure artificial neural network model calibration sets R, when with
When checking collection sample is tested, it is 0.231 for 0.928, RMSEP to be verified collection model R.It can be seen that, in the 3 kinds of information set up
It is in transfer mode artificial neural network pattern, optimum with Jordan-Elman nets structural models, and standard nets knots
Structure model and jump connection nets structural models predict the outcome and are closer to, and prediction effect is less desirable.
30 checking collection fresh leaf samples are predicted using optimal Jordan-Elman nets structural models, differentiate knot
Fruit is specifically shown in Table 3.From table 3 it can be seen that optimal calibration set model can accurately predict the fresh leaf sample of unknown credit rating,
Ideal prediction effect is reached, has differentiated that accuracy rate is 100%.It can be seen that, Jordan-Elman nets structure artificials nerve
Network model can realize quick, lossless, the accurate differentiation of different quality grade fresh leaf sample.
3 30 checking collection samples of table predict the outcome
Present invention application near-infrared spectrum technique, first scanning obtain the near infrared spectrum of fresh leaf sample, effectively reduce noise
After information, principal component analysiss are carried out to sample spectra and tries to achieve Score1 values and Score2 values, but cannot distinguish between different quality grade
Fresh leaf sample;Then using the Artificial Neural Network with good nonlinear characteristic, 3 main constituents in the past are input
Value, establishes standard nets, the difference of jump connection nets and Jordan-Elman tri- kinds of structures of nets
The fresh leaf artificial nerve network model of credit rating, obtains optimal with Jordan-Elman nets structural models prediction effect, tests
Card collection model R is 0.231 for 0.928, RMSEP.Quick, the accurate differentiation of different quality grade tea fresh leaves is realized, effectively
Solve the difficult problem that different quality grade tea fresh leaves sample judges to exist.Meanwhile, result of study is also carried for the stay in grade of Folium Camelliae sinensis
A kind of safeguard of science is supplied.
Above content is with reference to specific preferred implementation further description made for the present invention, it is impossible to assert
The present invention be embodied as be confined to these explanations.For general technical staff of the technical field of the invention,
On the premise of without departing from present inventive concept, some simple deduction or replace can also be made, said structure should all be considered as belonging to
Protection scope of the present invention.
Claims (5)
1. a kind of method that near infrared spectrum quickly judges different quality grade tea fresh leaves, it is characterised in that scanning obtains different
The fresh leaf sample near infrared spectrum of credit rating, then carries out principal component analysiss to fresh leaf sample spectra, then with main constituent as defeated
The Artificial Neural Network Prediction Model for entering the different quality grade fresh leaf that value sets up much information transfer mode judges fresh leaf not
Homogenous quantities grade, specifically includes following steps:
Step one, fresh leaf sample collecting and classification
Gather two kinds of different quality grades tea fresh leaves sample of the same race respectively, according to the difference of credit rating, by tea fresh leaves sample with
Machine is divided into 2 set of calibration set and checking collection;
Step 2, spectra collection
The near infrared spectrum for obtaining whole fresh leaf samples is scanned using Fourier-type near infrared spectrometer;
Step 3, Pretreated spectra
Applied Chemometrics software carries out derivation and smooth pretreatment to the near infrared spectrum of whole fresh leaf samples, then will be fresh
Leaf sample spectra is converted into paired data point;
Step 4, fresh leaf sample spectra principal component analysiss
Principal component analysiss are carried out to the spectroscopic data of whole fresh leaf samples using Matlab softwares, whole fresh leaf sample spectras are tried to achieve
The score Score1 value and Score2 values, number of principal components and its contribution rate of data;
Step 5, set up Artificial Neural Network Prediction Model
Front 3 main constituents with calibration set sample spectra, are passed through as input value with different quality grade fresh leaf type as output valve
Optimize repeatedly, set up standard nets, tri- kinds of information of jump connection nets and Jordan-Elman nets and pass
Mode fresh leaf different quality grade Artificial Neural Network Prediction Model is passed, compares three kinds of model coefficient R and validation-cross
Root-mean-square variance RMSECV value,
Wherein coefficient R formula is:
Validation-cross root-mean-square variance RMSECV formula is:
In formula, R is correlation coefficient, and n represents sample number, yiAnd yi' it is respectively the credit rating actual measurement of i-th sample in sample sets
Value and credit rating predictive value,For the meansigma methodss of the measured value of i-th sample in sample sets, i≤n in formula,
Wherein as best model, Jing compares the model of and validation-cross root-mean-square variance RMSECV minimum maximum with coefficient R
After obtain optimal calibration set model;
Step 6, model checking
To avoid the occurrence of overfitting phenomenon, application verification collection sample is examined to the three kinds of calibration set forecast result of model for obtaining
Test, acquired results coefficient R and checking collection mean square deviation RMSEP represent that wherein coefficient R is bigger and verifies collection mean square deviation
RMSEP is more little then to represent that test effect is better, if the credit rating predictive value of the near infrared spectrum for now obtaining and credit rating
Measured value is basically identical, then it represents that fine to the prediction effect of checking collection sample, and optimal calibration set model accurately can be predicted
The different quality grade of fresh leaf sample,
Wherein checking collection mean square deviation RMSEP formula is:
In formula, n represents sample number, yiAnd yi' it is respectively the credit rating measured value and credit rating of i-th sample in sample sets
Predictive value, i≤n in formula.
2. the method that a kind of near infrared spectrum according to claim 1 quickly judges different quality grade tea fresh leaves, which is special
Levy and be:In described step one, fresh leaf sample size is 100 parts, wherein the superfine first-class of tea fresh leaves sample and superfine second-class each 50
Individual, fresh leaf sample is according to 7:3 ratio random division is calibration set and checking collection.
3. the method that a kind of near infrared spectrum according to claim 1 quickly judges different quality grade tea fresh leaves, which is special
Levy and be:The fresh leaf sample plucked in described step one is to open up at the beginning of bud, one leaf of a bud.
4. the method that a kind of near infrared spectrum according to claim 1 quickly judges different quality grade tea fresh leaves, which is special
Levy and be:Fourier-type near infrared spectrometer in the step 2 is with silent winged your II type Fu of Antaris of generation of U.S.'s match
Leaf near infrared spectrometer, spectral scanning range 4000-10000cm-1, resolution 8cm-1, detector is InGaAs, each sample
10 spectrum of collection, every time scanning 64 times take the final spectrum of the meansigma methodss as the sample of 10 collection spectrum.
5. the method that a kind of near infrared spectrum according to claim 1 quickly judges different quality grade tea fresh leaves, which is special
Levy and be:Chemo metric software in the step 3 is 7.0 software of TQ Analyst 9.4.45 softwares and OPUS.
Priority Applications (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN201610930404.XA CN106568738A (en) | 2016-10-31 | 2016-10-31 | Method of using near infrared spectroscopy to rapidly determine fresh leaves of tea in different quality grades |
Applications Claiming Priority (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN201610930404.XA CN106568738A (en) | 2016-10-31 | 2016-10-31 | Method of using near infrared spectroscopy to rapidly determine fresh leaves of tea in different quality grades |
Publications (1)
Publication Number | Publication Date |
---|---|
CN106568738A true CN106568738A (en) | 2017-04-19 |
Family
ID=58534222
Family Applications (1)
Application Number | Title | Priority Date | Filing Date |
---|---|---|---|
CN201610930404.XA Pending CN106568738A (en) | 2016-10-31 | 2016-10-31 | Method of using near infrared spectroscopy to rapidly determine fresh leaves of tea in different quality grades |
Country Status (1)
Country | Link |
---|---|
CN (1) | CN106568738A (en) |
Cited By (12)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN107271396A (en) * | 2017-08-14 | 2017-10-20 | 扬州大学 | The quick determination method of general flavone content in a kind of tealeaves |
CN108931483A (en) * | 2017-05-25 | 2018-12-04 | 中科遥感(深圳)卫星应用创新研究院有限公司 | Agricultural product compose line library and inquiry system |
CN108960844A (en) * | 2017-05-25 | 2018-12-07 | 中科遥感(深圳)卫星应用创新研究院有限公司 | Spectrum line is traced to the source platform |
CN109086254A (en) * | 2017-12-29 | 2018-12-25 | 东北电力大学 | It is evaluated based on hyperspectral technique paraffin internal composition integrated level |
CN109297929A (en) * | 2018-11-29 | 2019-02-01 | 国药集团同济堂(贵州)制药有限公司 | A method of salvia piece quality grading is established using near infrared technology |
CN110320174A (en) * | 2019-06-14 | 2019-10-11 | 湖北省农业科学院果树茶叶研究所 | Using the method for polynomial net structure artificial neural network quick predict Yuanan yellow tea bored yellow time |
CN110320173A (en) * | 2019-06-14 | 2019-10-11 | 湖北省农业科学院果树茶叶研究所 | The method for rapidly judging of machine fresh tea picking mee tea vehicle tinctorial pattern product grade based on particle swarm optimization algorithm |
CN110334714A (en) * | 2019-06-14 | 2019-10-15 | 湖北省农业科学院果树茶叶研究所 | A kind of machine based on artificial neural network technology adopts mee tea vehicle tinctorial pattern product grade prediction technique |
CN110334174A (en) * | 2019-07-15 | 2019-10-15 | 武汉中海庭数据技术有限公司 | A kind of POS data method for correcting error and system |
CN110361334A (en) * | 2019-06-14 | 2019-10-22 | 湖北省农业科学院果树茶叶研究所 | The method for adopting mee tea vehicle tinctorial pattern product grade using general regression structure non-destructive prediction machine |
CN111842191A (en) * | 2020-08-03 | 2020-10-30 | 湖北省农业科学院果树茶叶研究所 | Sorting mechanism for distinguishing grades of fresh tea leaves |
CN113552089A (en) * | 2021-07-06 | 2021-10-26 | 中国农业科学院茶叶研究所 | Digital nondestructive testing method for black tea withering degree |
-
2016
- 2016-10-31 CN CN201610930404.XA patent/CN106568738A/en active Pending
Non-Patent Citations (5)
Title |
---|
IRENA IŠTOKA OTKOVIC′ ET AL.: "Calibration of microsimulation traffic model using neural network approach", 《EXPERT SYSTEMS WITH APPLICATIONS》 * |
QINGHUA ZHANG ET AL.: ""Prediction of Performance Indices and Optimal Parameters of Rough Rice Drying using Neural Networks"", 《BIOSYSTEMS ENGINEERING》 * |
王胜鹏 等: "近红外光谱结合神经网络判别茶鲜叶产地", 《中国计量学院学报》 * |
赵杰文 等: "《茶叶质量与安全检测技术及分析方法》", 31 March 2011, 中国轻工业出版社 * |
陈兰珍 等: "《蜂蜜近红外光谱检测技术》", 31 January 2012, 中国轻工业出版社 * |
Cited By (13)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN108931483A (en) * | 2017-05-25 | 2018-12-04 | 中科遥感(深圳)卫星应用创新研究院有限公司 | Agricultural product compose line library and inquiry system |
CN108960844A (en) * | 2017-05-25 | 2018-12-07 | 中科遥感(深圳)卫星应用创新研究院有限公司 | Spectrum line is traced to the source platform |
CN107271396A (en) * | 2017-08-14 | 2017-10-20 | 扬州大学 | The quick determination method of general flavone content in a kind of tealeaves |
CN109086254A (en) * | 2017-12-29 | 2018-12-25 | 东北电力大学 | It is evaluated based on hyperspectral technique paraffin internal composition integrated level |
CN109297929A (en) * | 2018-11-29 | 2019-02-01 | 国药集团同济堂(贵州)制药有限公司 | A method of salvia piece quality grading is established using near infrared technology |
CN110320173A (en) * | 2019-06-14 | 2019-10-11 | 湖北省农业科学院果树茶叶研究所 | The method for rapidly judging of machine fresh tea picking mee tea vehicle tinctorial pattern product grade based on particle swarm optimization algorithm |
CN110320174A (en) * | 2019-06-14 | 2019-10-11 | 湖北省农业科学院果树茶叶研究所 | Using the method for polynomial net structure artificial neural network quick predict Yuanan yellow tea bored yellow time |
CN110334714A (en) * | 2019-06-14 | 2019-10-15 | 湖北省农业科学院果树茶叶研究所 | A kind of machine based on artificial neural network technology adopts mee tea vehicle tinctorial pattern product grade prediction technique |
CN110361334A (en) * | 2019-06-14 | 2019-10-22 | 湖北省农业科学院果树茶叶研究所 | The method for adopting mee tea vehicle tinctorial pattern product grade using general regression structure non-destructive prediction machine |
CN110334174A (en) * | 2019-07-15 | 2019-10-15 | 武汉中海庭数据技术有限公司 | A kind of POS data method for correcting error and system |
CN110334174B (en) * | 2019-07-15 | 2021-04-09 | 武汉中海庭数据技术有限公司 | POS data deviation rectifying method and system |
CN111842191A (en) * | 2020-08-03 | 2020-10-30 | 湖北省农业科学院果树茶叶研究所 | Sorting mechanism for distinguishing grades of fresh tea leaves |
CN113552089A (en) * | 2021-07-06 | 2021-10-26 | 中国农业科学院茶叶研究所 | Digital nondestructive testing method for black tea withering degree |
Similar Documents
Publication | Publication Date | Title |
---|---|---|
CN106568738A (en) | Method of using near infrared spectroscopy to rapidly determine fresh leaves of tea in different quality grades | |
CN105928901B (en) | A kind of near-infrared quantitative model construction method that qualitative, quantitative combines | |
CN108872132A (en) | A method of fresh tea leaves kind is differentiated using near infrared spectrum | |
CN104062257A (en) | Method for determining total flavone content of solution based on near infrared spectroscopy | |
CN104062258B (en) | Method for rapid determination of soluble solids in compound ass-hide glue pulp by near infrared spectroscopy | |
CN108844917A (en) | A kind of Near Infrared Spectroscopy Data Analysis based on significance tests and Partial Least Squares | |
CN105181650B (en) | A method of quickly differentiating local tea variety using near-infrared spectrum technique | |
CN104062259A (en) | Method for rapid determination of total saponin content in compound ass-hide glue pulp by near infrared spectroscopy | |
CN112414967B (en) | Near infrared quality control method for rapidly detecting processing of cattail pollen charcoal in real time | |
Yun et al. | Identification of tea based on CARS‐SWR variable optimization of visible/near‐infrared spectrum | |
CN110186871A (en) | A kind of method of discrimination in the fresh tea leaves place of production | |
CN110308111B (en) | Method for rapidly predicting time for smoldering yellow tea by using near infrared spectrum technology | |
CN110084227A (en) | Mode identification method based on near-infrared spectrum technique | |
CN106568740A (en) | Method for rapid judging of varieties of fresh tea leaves by near infrared spectroscopy | |
CN106442400A (en) | Method for rapidly discriminating fresh tea leaves from different soil types through near infrared spectra | |
CN113310934A (en) | Method for quickly identifying milk cow milk mixed in camel milk and mixing proportion thereof | |
CN106442399A (en) | Method for distinguishing fresh leaves of same variety of tea from different cultivation environments by aid of near infrared spectra | |
CN110186870B (en) | Method for distinguishing fresh tea leaf producing area of Enshi Yulu tea by extreme learning machine spectrum model | |
CN106568739A (en) | Method for rapidly determining fresh tea leaves at different altitudes through near infrared spectroscopy | |
Abdullah et al. | ANN Diagnostic System for Various Grades of Yellow Flesh Watermelon based on the Visible light and NIR properties | |
CN109685099A (en) | A kind of apple variety discriminating conduct of the preferred fuzzy clustering of spectral band | |
CN106568741A (en) | Method used for determining producing areas of tea fresh leaves via near infrared spectrum rapidly | |
EP3961189B1 (en) | Cigarette mainstream smoke spectral data sensory evaluation method | |
CN112881323B (en) | Quality evaluation method of cigarette mainstream smoke | |
CN110320173A (en) | The method for rapidly judging of machine fresh tea picking mee tea vehicle tinctorial pattern product grade based on particle swarm optimization 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 | ||
RJ01 | Rejection of invention patent application after publication |
Application publication date: 20170419 |