CN107024450A - A kind of method for differentiating different brands and hop count milk powder based on near-infrared spectrum technique - Google Patents
A kind of method for differentiating different brands and hop count milk powder based on near-infrared spectrum technique Download PDFInfo
- Publication number
- CN107024450A CN107024450A CN201710189560.XA CN201710189560A CN107024450A CN 107024450 A CN107024450 A CN 107024450A CN 201710189560 A CN201710189560 A CN 201710189560A CN 107024450 A CN107024450 A CN 107024450A
- Authority
- CN
- China
- Prior art keywords
- milk powder
- sample
- class
- hop count
- spectrum
- 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
- 235000013336 milk Nutrition 0.000 title claims abstract description 89
- 239000008267 milk Substances 0.000 title claims abstract description 89
- 210000004080 milk Anatomy 0.000 title claims abstract description 89
- 239000000843 powder Substances 0.000 title claims abstract description 87
- 238000000034 method Methods 0.000 title claims abstract description 43
- 238000002329 infrared spectrum Methods 0.000 title claims abstract description 13
- 238000001228 spectrum Methods 0.000 claims abstract description 44
- 230000003595 spectral effect Effects 0.000 claims abstract description 9
- 238000012549 training Methods 0.000 claims description 17
- 235000008476 powdered milk Nutrition 0.000 claims description 9
- 230000008859 change Effects 0.000 claims description 6
- 230000008569 process Effects 0.000 claims description 6
- 238000012937 correction Methods 0.000 claims description 5
- 230000005856 abnormality Effects 0.000 claims description 3
- 238000011985 exploratory data analysis Methods 0.000 claims description 3
- 230000002159 abnormal effect Effects 0.000 claims description 2
- 235000013399 edible fruits Nutrition 0.000 claims description 2
- 238000004445 quantitative analysis Methods 0.000 claims description 2
- 238000005457 optimization Methods 0.000 claims 1
- 238000012360 testing method Methods 0.000 abstract description 11
- 238000012850 discrimination method Methods 0.000 abstract description 4
- 238000011109 contamination Methods 0.000 abstract description 2
- 239000011159 matrix material Substances 0.000 description 11
- 238000000513 principal component analysis Methods 0.000 description 11
- 238000004458 analytical method Methods 0.000 description 7
- 238000010586 diagram Methods 0.000 description 7
- 239000013598 vector Substances 0.000 description 7
- 238000005516 engineering process Methods 0.000 description 6
- 235000013350 formula milk Nutrition 0.000 description 6
- 238000001514 detection method Methods 0.000 description 4
- 238000004519 manufacturing process Methods 0.000 description 4
- 239000000463 material Substances 0.000 description 4
- 230000000694 effects Effects 0.000 description 3
- 238000007689 inspection Methods 0.000 description 3
- 229920000877 Melamine resin Polymers 0.000 description 2
- 238000009825 accumulation Methods 0.000 description 2
- 235000013365 dairy product Nutrition 0.000 description 2
- JDSHMPZPIAZGSV-UHFFFAOYSA-N melamine Chemical compound NC1=NC(N)=NC(N)=N1 JDSHMPZPIAZGSV-UHFFFAOYSA-N 0.000 description 2
- 235000016709 nutrition Nutrition 0.000 description 2
- 238000005070 sampling Methods 0.000 description 2
- 241000894007 species Species 0.000 description 2
- 102000009091 Amyloidogenic Proteins Human genes 0.000 description 1
- 108010048112 Amyloidogenic Proteins Proteins 0.000 description 1
- 108010073771 Soybean Proteins Proteins 0.000 description 1
- 238000013459 approach Methods 0.000 description 1
- 230000009286 beneficial effect Effects 0.000 description 1
- 238000010668 complexation reaction Methods 0.000 description 1
- 239000002131 composite material Substances 0.000 description 1
- 230000006835 compression Effects 0.000 description 1
- 238000007906 compression Methods 0.000 description 1
- 238000012790 confirmation Methods 0.000 description 1
- 239000000470 constituent Substances 0.000 description 1
- 230000007423 decrease Effects 0.000 description 1
- 238000009795 derivation Methods 0.000 description 1
- 238000013461 design Methods 0.000 description 1
- 238000009826 distribution Methods 0.000 description 1
- 238000000605 extraction Methods 0.000 description 1
- 238000001914 filtration Methods 0.000 description 1
- 235000021393 food security Nutrition 0.000 description 1
- 230000036541 health Effects 0.000 description 1
- 230000008676 import Effects 0.000 description 1
- 230000006872 improvement Effects 0.000 description 1
- 239000004615 ingredient Substances 0.000 description 1
- 238000012417 linear regression Methods 0.000 description 1
- 238000013507 mapping Methods 0.000 description 1
- 238000005259 measurement Methods 0.000 description 1
- 238000012986 modification Methods 0.000 description 1
- 230000004048 modification Effects 0.000 description 1
- 238000000491 multivariate analysis Methods 0.000 description 1
- 230000035764 nutrition Effects 0.000 description 1
- 238000003359 percent control normalization Methods 0.000 description 1
- 238000002360 preparation method Methods 0.000 description 1
- 238000002203 pretreatment Methods 0.000 description 1
- 238000002331 protein detection Methods 0.000 description 1
- 238000012797 qualification Methods 0.000 description 1
- 230000003252 repetitive effect Effects 0.000 description 1
- 230000004044 response Effects 0.000 description 1
- 230000035939 shock Effects 0.000 description 1
- 235000019710 soybean protein Nutrition 0.000 description 1
- 238000004611 spectroscopical analysis Methods 0.000 description 1
- 238000010183 spectrum analysis Methods 0.000 description 1
- 230000006641 stabilisation Effects 0.000 description 1
- 238000011105 stabilization Methods 0.000 description 1
- 238000012546 transfer Methods 0.000 description 1
- 230000017105 transposition Effects 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
- 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 kind of method for differentiating different brands and hop count milk powder based on near-infrared spectrum technique, it is related to milk powder authentication technique field.Discrimination method is：Near-infrared original spectrum is gathered, and it is modeling wave band or wavelength to choose the obvious spectral band of feature or wavelength in original spectrum wave band；Pretreated spectra, the foundation of PLS DA disaggregated models, the checking of PLS DA disaggregated models, the foundation of principal component class model, the checking of PCA class models, the discriminating of milk powder brand and hop count.This method is the discrimination method of a kind of easy, quick, lossless free of contamination milk powder brand and hop count, utilize near infrared spectrometer, quickly unknown milk powder brand and hop count can be differentiated without professional testing staff, it can be promoted in milk powder intermediate links, effective supervision to quality of milk powder is realized, prevents illegal retailer from producing and selling the phenomenons such as personation, inferior milk power and occurring.
Description
Technical field：
The present invention relates to a kind of method for differentiating different brands and hop count milk powder based on near-infrared spectrum technique, belong to milk powder
Authentication technique field.
Background technology：
Milk powder, as important dairy products, is the important nutrition intake approach of infant.As the future of motherland, society is each
The obligated abundant food security and quality for ensureing infant in boundary.The safety and quality of milk powder decide that the body of infant is good for
Health.
In recent years, milk powder safety accident continuously emerges, and " Sanlu milk powder " event order society of 2008 shocks and distressed,
Shade has been coverd with to China's dairy industry.Sight is gradually turned to foreign brand by father and mother.Foreign brand is due to its milk supply high-quality, matter
Amount stabilization is gradually by the favor of father and mother.In recent years, as Hai Tao, extra large generation develop rapidly, to drink foreign milk powder turn into safety
Reliable synonym.In April, 2016, the Shanghai police have cracked a personation milk powder case, discover and seize the tank of personation Abbott Laboratories milk powder 1.7 ten thousand.
At the same time, Shi Yaojian general bureaus respond in official website and claimed, and the police test to the counterfeit milk powder discovered and seized, and product meets national mark
Standard, and in the absence of security risk.Current event is that criminal will sell after qualified milk powder transducer package, illegal with what is backspreaded
Behavior.Domestic brand is changed into import brand by they, and minor brands are changed into famous brand.Although not causing the damage of infant's body
Wound, but economic loss is brought to consumer, also encroach on by the interests of infringement enterprise.
Near-infrared spectrum technique is the analysis and detection technology grown up 1950s, with quick, lossless, green
Color, it is safe the features such as be used widely in every profession and trade rapidly.In recent years, lot of domestic and international scholar attempts to apply near infrared technology
Detect nutritional ingredient, risk indicator or the adulterated material of milk powder in milk powder.For example：Chang Min etc. is using the inspection of near-infrared diffusing reflection spectrum
Survey protein content in milk powder.By wavelength preferably with data prediction method choice, establish and quick determine variety classes milk
The near-infrared quick determination method of amyloid proteins matter content.Whether Qian Feng is established and is contained in measure milk powder by near infrared technology
The forecast model of melamine is added, preferable effect is achieved.Tang Yulian application linear discriminants quick nondestructive, which differentiates, to be combined closely
Infrared spectrum differentiates infant and Milk powder for middle-aged and old people, two kinds of different age group milk of person in middle and old age and baby milk powder to different brands
Powder kind is judged that the prediction relative deviation to unknown sample is less than 5%, and discrimination reaches 100%.
In example above, by near-infrared spectrum technique be applied to quality of milk powder detection and judgement in, but exist it is following not
Foot:1st, quality of milk powder is a complicated composite target, only by detecting one or more of active constituent contents in milk powder, no
Quality of milk powder can be reflected comprehensively.2nd, the exogenous material species added in milk powder is various, illegal retailer's addition soybean protein, paste
The materials such as essence, melamine, malt extract are to improve protein detection content in milk powder, and to reach adulterated milk powder, milk powder product is examined
The accurate purpose of mark.But such method needs voluntarily to prepare adulterated sample when setting up model or with actually adulterated sample as building
Apperance sheet, operation difficulty is big, it is impossible to meet the demand of the identification adulterated material of variety classes.3rd, by setting up true and false milk powder or specific
The near-infrared discrimination model of the milk powder of brand, can recognize different brands milk powder.But different hop counts for same brand or not
Milk powder with the place of production differentiates there is object limitation, it is impossible to meet the requirement of the omnifarious fraud situation of strike.4th, milk powder brand
It is various with hop count species complexity, version, differentiate only by a kind of method and just draw a conclusion, might have the production of the phenomenon of erroneous judgement
It is raw,
Offset minimum binary-discriminant analysis (PLS-DA) method is a kind of linear regression method for having a supervision, with compression number
It is the discriminant analysis side based on PLS using binary class variable as Y variables according to the premium properties of simultaneously characteristic information extraction
Method.Companion matrix is take into account in structural factor class members's information is provided with code form, therefore with efficient discriminating
Ability.
PLS-DA is a kind of mode identification method for having supervision based on offset minimum binary, and it utilizes priori classification knowledge,
A PLS-DA disaggregated model is set up, the ownership of such Model checking testing sample is then utilized.Its main process is as follows：
The appropriate known class sample spectra of selection is used as training set.In training set, to the spectrum of any one sample
xiCan be with vector representation：
xi=[xi1 xi2 … xin]
The sample sets constituted to m sample, may make up the spectrum data matrix of m × n rank：
It is response variable by the category setting of m sample, and is quantified, this class value is 1, and non-class value is 0, this
Sample, it is possible to obtain the classification matrix Y of the rank of m × 1m×1。
By PLS, the offset minimum binary class model of spectrum matrix and classification matrix is set up：
Ym×1=Xm×nBn×1
In formula, Bn×1For regression coefficient matrix.
For testing sample xnew, the class label of testing sample can be calculated according to following formula：
ynew=xnewBn×1
To ynewAfter being rounded, according to its size, the classification of testing sample can be accurately differentiated.Decision rule is：
1. y is worked asnew＞ 0.5, and during deviation ＜ 0.5, judge that sample belongs to this class；2. y is worked asnew＜ 0.5, during deviation ＜ 0.5, judges sample
It is not belonging to this class；3. y is worked asnewDeviation >=0.5, judge it is unstable.
The general handling process of PLS-DA sorting algorithms is：
(1) spectrum pre-treatment, such as derivation, centralization, to eliminate the effects of the act are carried out to spectroscopic data；
(2) application PLS is modeled；
(3) rejecting abnormalities sample, repetitive operation, interactivity checking, determines the suitable number of principal components of class model；
(4) testing sample is predicted using class model；
(5) according to predicting the outcome, classification is judged using decision rule.
Such a algorithmic procedure is complex, and qualification result has certain error.
Hotelling T2What statistic was characterized is that one kind of principal component model interior change is estimated.Pass through principal component mould
The fluctuation of principal component vector mould inside type reflects the situation of multivariable change, and it represents the average and sampling of multivariate data
Distance of the data point in principal component plane between mapping point.If T2Statistic is limited beyond control, then illustrates its sample quality
There is exception, otherwise sample quality is normal.Hotelling T2Statistic has various Optimalities, and it is consistent most strong
Invariant test, with Optimalities such as permission property and the properties of minimization greater risk.
Using near infrared spectrometer, the near infrared spectrum data of separate sources sample is gathered, you can constitute the spectrum of sample
Data set.Any one sample standard deviation gathered can be characterized with one group of parameter, these parameters are exactly original spectrum in fact
Data, can be with vector representation：xi=[xi1 xi2 。。。 xin].So, the sample set constituted to m sample, so that it may be expressed as
Spectrum data matrix Xm × n of m × n ranks, in matrix Xm × n, the both information comprising sample properties, also comprising sample characteristics or
The information of variable.From the point of view of mathematical angle, the point set in the n dimensional feature spaces that sample set can be made up of characteristic coordinates axle is come table
Show.One sample one point of correspondence, n feature contains the full detail of sample properties, relation, change between sample and sample
Relation between relation and sample and variable between amount and variable etc., is exactly that 3 to be processed of multi-variables analysis is asked substantially
Topic.In multi-variables analysis, the class model of different classes of sample is set up using principal component analysis, can be characterized in sample set sample with
Implication relation between sample, between sample and class.(1) is divided spectrum data matrix as the following formula for principal component analysis (PCA)
Solution,
Ti represents Xm × n score vector in formula；Pi represents Xm × n load vectors；A is factor number；T representing matrixs
Transposition；E represents Xm × n error term.
T is carried out to matrix Xm × n2Examine, T2Inspection is a kind of conventional multivariable method of inspection, is a kind of multivariate analysis
In significance test to average difference, it is assumed that H0∶μ1=μ2=...=μp, calculate respectively each variable average constitute it is equal
Vector and covariance matrix V, then Hotelling T2StatisticRespectively
The Hotelling T that equal vector compares2Statistic has following relation with statistic F:
Judged by F distributions.
During actual sample kind judging, such a method carries out class prediction to sample, there is certain limitation
Property.For border sample, it can decline Model Identification ability because various factors influence, causes the erroneous judgement of certain probability.
Therefore need to find a kind of safe and reliable, the accurate and simple and quick milk of measurement result at this stage at present
The detection method of powder brand and hop count.
The content of the invention：
In view of the above-mentioned problems, the technical problem to be solved in the present invention is to provide it is a kind of easy, quick, utilize near infrared spectrum
Instrument, quickly mirror method for distinguishing can be carried out without professional testing staff to unknown milk powder brand and hop count.
The present invention's differentiates that the method for different brands and hop count milk powder is based on near-infrared spectrum technique：Step 1: near red
Outer original spectra collection：The qualified milk powder of certified products of different brands and hop count is collected, under the conditions of constant temperature and humidity, near infrared light is utilized
Spectrometer, gathers the near-infrared diffusing reflection spectrum of different type milk powder respectively, and spectra collection scope is：Wave number is in 12000cm-1-4000cm-1Between；
Step 2: it is modeling wave band or wavelength to choose the obvious spectral band of feature or wavelength in original spectrum wave band；
Step 3: Pretreated spectra：To modeling wave band or wavelength, using multiplicative scatter correction, standard normal change, micro-
Point, the one or more in the method such as smooth spectrum is pre-processed, to remove spectral noise, raising signal to noise ratio reaches reason
The modeling effect thought；
Step 4: the foundation of PLS-DA disaggregated models:By pretreated sample spectra, category PCA
Exploratory analysis is carried out to each class powdered milk sample spectrum, it is preliminary to reject spectral singularity sample；Each class sample is divided into training
Collection and checking collection, for the selection of training set and checking collection, with reference to GB/T 29858《The quantitative analysis of molecular spectrum Multivariate Correction is led to
Then》, disaggregated model is set up with training set sample, in modeling process, by continuing to optimize model, the discriminating power of model is improved
And robustness, abnormal sample is identified, suitable number of principal components modeling is selected；
Step 5: the checking of PLS-DA disaggregated models:Using verifying that the disaggregated model that set pair is set up is verified, by knowing
Rate does not assess the discriminating power of class model；
Step 6: the foundation of principal component class model：Spectrum is handled suddenly step by step by the 4th, after rejecting abnormalities sample,
Divide training set and checking collects, the PCA class models of each class powdered milk sample are set up respectively with training set sample, in modeling process
Continue to optimize, choose optimum number of principal components modeling, pass through Hotelling T2Statistic monitors unknown sample, with
Hotelling T299% confidence limit of statistic is limited as control, and 95% confidence limit is limited as early warning, the class model energy of foundation
Whether confirmation unknown sample belongs to the category；
Step 7: the checking of PCA class models：By the milk powder principal component class model of foundation, Hotelling T are extracted2System
Metering is verified to different type milk powder sample；
Step 8: the discriminating of milk powder brand and hop count:By the PLS-DA model applications of the milk powder brand established and hop count
In unknown powdered milk sample, sample class value is calculated, the brand and hop count of milk powder are first judged according to class label, recycles PLS-DA to sentence
Determine result, transfer the corresponding principal component class model of the category, extract Hotelling T2Statistic judged, such as two methods
Judgement belongs to the category, then the sample is certified products.
Beneficial effects of the present invention：This method by gathering the milk powder near-infrared diffusing reflection spectrums of different brands and hop count,
By being pre-processed to spectrum, discrimination model is set up using chemometrics method, using verifying that the set pair analysis model validity enters
Row checking；By the near infrared spectrum input model of unknown powdered milk sample, class label of the sample between inhomogeneity is calculated, according to sentencing
Then which kind of milk powder preliminary judgement unknown sample belongs to set pattern.Again by the category principal component class model of preliminary judgement, extract
Hotelling T2Statistic is verified that this method is a kind of easy, quick, lossless free of contamination milk powder brand and hop count
Discrimination method, using near infrared spectrometer, quickly can reflect without professional testing staff to unknown milk powder brand and hop count
Not, it can be promoted in milk powder intermediate links, realize effective supervision to quality of milk powder, prevent illegal retailer from producing and selling false
Emit, the phenomenon such as inferior milk power occurs.
Brief description of the drawings：
For ease of explanation, the present invention is described in detail by following specific implementations and accompanying drawing.
Fig. 1 is the flow chart of the detection method of the present invention；
Fig. 2 sets up the statistical results chart of PLS-DA models for different number of principal components in the embodiment of the present invention 1；
Fig. 3 sets up the statistics knot of PCA class models for the different number of principal components of the class sample of A, B, C, D tetra- in the embodiment of the present invention 1
Fruit is schemed；
Fig. 4 is the principal component scores schematic diagram of PLS-DA models first, second of 4 kinds of milk powder in the embodiment of the present invention 1.
Fig. 5 is the principal component scores schematic diagram of PCA class models first, second of A milk powder in the embodiment of the present invention 1.
Fig. 6 verifies the identification result schematic diagram collected for the PLS-DA models set up in the embodiment of the present invention 1 to A classes sample.
Fig. 7 verifies the identification result schematic diagram collected for the PLS-DA models set up in the embodiment of the present invention 1 to B classes sample.
Fig. 8 verifies the identification result schematic diagram collected for the PLS-DA models set up in the embodiment of the present invention 1 to C classes sample.
Fig. 9 verifies the identification result schematic diagram collected for the PLS-DA models set up in the embodiment of the present invention 1 to D classes sample.
Figure 10 is predict the outcome schematic diagram of the class model to four kinds of milk powder of A classes milk powder foundation in the embodiment of the present invention 1.
Embodiment：
Present embodiment is using following examples come described in further detail.
Embodiment 1, one kind as shown in Figure 1 is based on near-infrared spectral analysis technology discriminating different brands and hop count milk powder
Method, comprise the following steps：
Step one：The collection of milk powder near infrared spectrum：Using Fourier Transform Near Infrared instrument collection different brands and
The near-infrared diffusing reflection spectrum of hop count milk powder, detailed step is as follows：
(1-1) preparation of samples：(A milk powder is that Holland produces, B milk powder is moral to the baby milk powder of 4 kinds of 1 section of different brands of purchase
Domestic, C milk powder is that New Zealand produces, D milk powder is German production), tetra- kinds of milk powder of A, B, C, D buy difference respectively in different regular channels
50, the sample of date of manufacture, altogether 200.
Powdered milk sample is fitted into special sampling cup by (1-2), is placed near infrared spectrometer and is acquired, utilizes computer
On spectra collection software obtain original spectrum.
(1-3) spectra collection：20 DEG C of laboratory temperature constant temperature, the design parameter of near infrared spectrometer：Spectra collection scope：
12000 wave number (cm-1) -4000 wave number (cm-1)；Resolution ratio 4cm-1, scanning times：64 times, each powdered milk sample is scanned 1 time.
Step 2：Choose modeling wave band：The spectral band of the original spectrum of acquisition, short wavelength regions information is less, selection
8500 wave number (cm-1) -4100 wave number (cm-1) it is used as modeling wave band.
Step 3：Pretreated spectra：Using multiplicative scatter correction, first differential, Savitzky-Golay filtering to original
Spectrum is handled.
Step 4：The foundation of PLS-DA models：Exploratory analysis is carried out to every class sample, spectral singularity sample is rejected, makes
With the spectrum treated through step 3：(4-1) will set each sample in the class sample of A, B, C, D tetra-, belong to this classification and set classification
For " 1 ", such sample class is not belonging to for " 0 ", and sample is divided into training set and checking collects, 30 milk powder per class milk powder are made
For training set sample, model is set up as training set for 120 altogether, remaining 80 sample checking collection；(4-2) determines model master
Component number, (R2X is used by the accumulation interpretability of modelcumRepresent) and the precision of prediction of model (interact validity with accumulative
Q2cumRepresent) determine to show the statistical results of different number of principal components modelings in the optimal number of principal components of model, Fig. 2, from Fig. 2
As can be seen that when number of principal components increases to 6, the contribution to lift scheme precision of prediction is not notable, so, model it is optimal
Number of principal components is defined as 5；(4-3) sets up PLS-DA models under the conditions of the optimal number of principal components of model, utilizes 120 samples of training set
Originally it is modeled, obtains the PLS-DA disaggregated models of 4 kinds of milk powder.
Step 5：The foundation of PCA class models：(5-1) sets up the PCA classes of all kinds of samples respectively to the class sample of A, B, C, D tetra-
Model.Per 30 samples in class milk powder as training set, remaining 20 are used as checking collection.(5-2) determines model number of principal components,
R2X (is used by the accumulation interpretability of modelcumRepresent) and the precision of prediction of model (interact validity Q2 with accumulativecumRepresent)
Come the system of the different number of principal components modeling that determines to show four classification PCA tear models in the optimal number of principal components of every class model, Fig. 3
Count result.By taking A class milk powder modules as an example, when number of principal components increases to 5, the 5th principal component newly increased predicts essence to lift scheme
The contribution of degree is not notable.So, the suitable number of principal components of A class milk powder class models should be 4.The like, B, C, D class sample class moulds
The suitable number of principal components of type is respectively 3,5,3.
Step 6：The judgement of milk powder kind：Checking 80 sample classification variate-values of collection are calculated, sample is judged according to decision rule
Originally which kind of belongs to.
Type decision is carried out to checking collection kind 4 class, 80 samples using the PLS-DA models of foundation, as a result such as accompanying drawing 6-
Shown in Fig. 9, the discriminating accuracy rate of the class sample of A, B, C, D tetra- checking collection is 100%.In the checking collection sample of accompanying drawing 6, A milk powder samples
This classified variable predicted value is all close to 1, and deviation is less than 0.5, and the classified variable predicted value of its excess-three class sample is close to 0, partially
Difference is less than 0.5.According to decision rule, the A class sample standard deviations that checking is concentrated are correctly validated, and other three classes samples are all identified as non-
This class is identical with accompanying drawing 8 to the analysis that predicts the outcome in accompanying drawing 7,8 and 9.Can obtain, the PLS-DA models of foundation to A, B,
The discriminating accuracy rate of the class milk powder sample of C, D tetra- is 100%.
By taking A models as an example, above result of determination is verified using the A class sample class models of foundation.As a result such as accompanying drawing
Shown in 10, the Hotelling T of A classes sample checking collection2Statistic is below 95% early warning limit.Its excess-three class sample checking collection
Hotelling T2Statistic is far above 99% control limit.Therefore, the discriminating accuracy rate of four class milk powder samples of foundation is
100%.
It is combined by both the above discrimination method, the erroneous judgement produced by a kind of limitation of method can be avoided.By mutual
Mutually verify, make identification result more accurately and reliably.
The general principle and principal character and advantages of the present invention of the present invention has been shown and described above.The technology of the industry
Personnel are it should be appreciated that the present invention is not limited to the above embodiments, and the simply explanation described in above-described embodiment and specification is originally
The principle of invention, without departing from the spirit and scope of the present invention, various changes and modifications of the present invention are possible, these changes
Change and improvement all fall within the protetion scope of the claimed invention.The claimed scope of the invention by appended claims and its
Equivalent thereof.
Claims (1)
1. a kind of method for differentiating different brands and hop count milk powder based on near-infrared spectrum technique, it is characterised in that：Specific mirror
Other method is：Step 1: near-infrared original spectrum is gathered：The qualified milk powder of certified products of different brands and hop count is collected, it is permanent in constant temperature
Under the conditions of wet, using near-infrared spectrometers, the near-infrared diffusing reflection spectrum of different type milk powder, spectra collection are gathered respectively
Scope is：Wave number is in 12000cm-1-4000cm-1Between；
Step 2: it is modeling wave band or wavelength to choose the obvious spectral band of feature or wavelength in original spectrum wave band；
Step 3: Pretreated spectra：To modeling wave band or wavelength, using multiplicative scatter correction, standard normal change, differential, put down
One or more in the method such as sliding are pre-processed to spectrum, to remove spectral noise, are improved signal to noise ratio, are reached and preferably build
Imitate fruit；
Step 4: the foundation of PLS-DA disaggregated models:By pretreated sample spectra, category PCA is to every
One class powdered milk sample spectrum carries out exploratory analysis, preliminary to reject spectral singularity sample；By each class sample be divided into training set and
Checking collection, for the selection of training set and checking collection, with reference to GB/T 29858《Molecular spectrum Multivariate Correction quantitative analysis general rule》,
Disaggregated model is set up with training set sample, in modeling process, by continuing to optimize model, the discriminating power of model is improved and steady
Strong property, abnormal sample is identified, and selects suitable number of principal components modeling；
Step 5: the checking of PLS-DA disaggregated models:Using verifying that the disaggregated model that set pair is set up is verified, pass through discrimination
Assess the discriminating power of class model；
Step 6: the foundation of principal component class model：Spectrum is handled suddenly step by step by the 4th, after rejecting abnormalities sample, divided
Training set and checking collect, and set up the PCA class models of each class powdered milk sample respectively with training set sample, in modeling process constantly
Optimization, chooses optimum number of principal components modeling, passes through Hotelling T2Statistic monitors unknown sample, with Hotelling
T299% confidence limit of statistic is limited as control, and 95% confidence limit is limited as early warning, and the class model of foundation can confirm unknown sample
Whether product belong to the category；
Step 7: the checking of PCA class models：By the milk powder principal component class model of foundation, Hotelling T are extracted2Statistic pair
Different type milk powder sample is verified；
Step 8: the discriminating of milk powder brand and hop count:The PLS-DA models of the milk powder brand established and hop count are applied to not
Know powdered milk sample, calculate sample class value, the brand and hop count of milk powder are first judged according to class label, recycles PLS-DA to judge knot
Really, the corresponding principal component class model of the category is transferred, Hotelling T are extracted2Statistic is judged that such as two methods judge
The category is belonged to, then the sample is certified products.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201710189560.XA CN107024450A (en) | 2017-03-27 | 2017-03-27 | A kind of method for differentiating different brands and hop count milk powder based on near-infrared spectrum technique |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201710189560.XA CN107024450A (en) | 2017-03-27 | 2017-03-27 | A kind of method for differentiating different brands and hop count milk powder based on near-infrared spectrum technique |
Publications (1)
| Publication Number | Publication Date |
|---|---|
| CN107024450A true CN107024450A (en) | 2017-08-08 |
Family
ID=59525464
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201710189560.XA Pending CN107024450A (en) | 2017-03-27 | 2017-03-27 | A kind of method for differentiating different brands and hop count milk powder based on near-infrared spectrum technique |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN107024450A (en) |
Cited By (12)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN107655852A (en) * | 2017-09-29 | 2018-02-02 | 广东出入境检验检疫局检验检疫技术中心 | The near infrared spectrum quick determination method of essential nutrient in baby formula milk powder |
| CN109030409A (en) * | 2018-08-30 | 2018-12-18 | 无锡迅杰光远科技有限公司 | A kind of Intellectual feeding bottle and milk power solution detection method |
| CN110907392A (en) * | 2019-11-15 | 2020-03-24 | 光钙(上海)高科技有限公司 | Melamine detection system based on infrared spectroscopic analysis and application and method thereof |
| CN112525850A (en) * | 2020-10-01 | 2021-03-19 | 华中农业大学 | Spectral fingerprint identification method for milk, mare, camel, goat and buffalo milk |
| CN112666112A (en) * | 2020-10-01 | 2021-04-16 | 华中农业大学 | Batch discrimination model and method for camel milk and mare milk |
| CN112666114A (en) * | 2020-10-01 | 2021-04-16 | 华中农业大学 | Method for identifying buffalo milk and mare milk by using spectrum |
| CN112666110A (en) * | 2020-10-01 | 2021-04-16 | 华中农业大学 | Spectral identification model and method for milk and goat milk |
| CN112666111A (en) * | 2020-10-01 | 2021-04-16 | 华中农业大学 | Method for quickly identifying milk and mare milk |
| CN112666113A (en) * | 2020-10-01 | 2021-04-16 | 华中农业大学 | Method for quickly identifying goat milk and mare milk |
| CN113049528A (en) * | 2021-02-02 | 2021-06-29 | 浙江理工大学 | Fiber component identification method and module based on near infrared spectrum |
| CN114780917A (en) * | 2022-06-20 | 2022-07-22 | 中国民航大学 | Milk powder doping non-directional detection method based on flow discrimination model |
| CN116165165A (en) * | 2023-04-25 | 2023-05-26 | 四川威斯派克科技有限公司 | Detection method for online real-time release of raw and auxiliary materials of medicines |
Citations (7)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN101995388A (en) * | 2009-08-26 | 2011-03-30 | 北京凯元盛世科技发展有限责任公司 | Near infrared quality control analysis method and system of tobacco |
| CN104297206A (en) * | 2014-11-04 | 2015-01-21 | 天津工业大学 | Quick liquid milk brand identifying method |
| CN104502299A (en) * | 2014-12-12 | 2015-04-08 | 深圳市计量质量检测研究院 | Method for identifying Wuchang rice-flower fragrance rice by virtue of near infrared spectrum technique |
| CN104713848A (en) * | 2015-03-27 | 2015-06-17 | 山东大学 | Method for distinguishing radix-paeoniae-alba production places on basis of near-infrared analysis technology |
| CN105606552A (en) * | 2016-02-04 | 2016-05-25 | 云南中烟工业有限责任公司 | Method for analyzing quality trends of cigarette tobaccos based on whole spectral range molecular spectrum |
| CN106124419A (en) * | 2016-06-22 | 2016-11-16 | 北京谱芸科技有限责任公司 | Milk powder type identifier device and milk powder type identifier system |
| CN108982692A (en) * | 2018-07-27 | 2018-12-11 | 深圳出入境检验检疫局食品检验检疫技术中心 | The method that elemental analysis-stable isotope mass spectrum differentiates the milk powder place of production |
-
2017
- 2017-03-27 CN CN201710189560.XA patent/CN107024450A/en active Pending
Patent Citations (7)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN101995388A (en) * | 2009-08-26 | 2011-03-30 | 北京凯元盛世科技发展有限责任公司 | Near infrared quality control analysis method and system of tobacco |
| CN104297206A (en) * | 2014-11-04 | 2015-01-21 | 天津工业大学 | Quick liquid milk brand identifying method |
| CN104502299A (en) * | 2014-12-12 | 2015-04-08 | 深圳市计量质量检测研究院 | Method for identifying Wuchang rice-flower fragrance rice by virtue of near infrared spectrum technique |
| CN104713848A (en) * | 2015-03-27 | 2015-06-17 | 山东大学 | Method for distinguishing radix-paeoniae-alba production places on basis of near-infrared analysis technology |
| CN105606552A (en) * | 2016-02-04 | 2016-05-25 | 云南中烟工业有限责任公司 | Method for analyzing quality trends of cigarette tobaccos based on whole spectral range molecular spectrum |
| CN106124419A (en) * | 2016-06-22 | 2016-11-16 | 北京谱芸科技有限责任公司 | Milk powder type identifier device and milk powder type identifier system |
| CN108982692A (en) * | 2018-07-27 | 2018-12-11 | 深圳出入境检验检疫局食品检验检疫技术中心 | The method that elemental analysis-stable isotope mass spectrum differentiates the milk powder place of production |
Non-Patent Citations (3)
| Title |
|---|
| D.WU 等: ""Short-Wave Near-Infrared Spectroscopy of Milk Powder for Brand Identification and Component Analysis"", 《J. DAIRY SCI》 * |
| 张初 等: ""近红外光谱技术用于豆浆粉品牌与假冒豆浆粉的鉴别"", 《光谱学与光谱分析》 * |
| 管骁 等: ""近红外光谱技术的乳粉品牌溯源研究"", 《光谱学与光谱分析》 * |
Cited By (13)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN107655852A (en) * | 2017-09-29 | 2018-02-02 | 广东出入境检验检疫局检验检疫技术中心 | The near infrared spectrum quick determination method of essential nutrient in baby formula milk powder |
| CN109030409A (en) * | 2018-08-30 | 2018-12-18 | 无锡迅杰光远科技有限公司 | A kind of Intellectual feeding bottle and milk power solution detection method |
| CN110907392A (en) * | 2019-11-15 | 2020-03-24 | 光钙(上海)高科技有限公司 | Melamine detection system based on infrared spectroscopic analysis and application and method thereof |
| CN112525850A (en) * | 2020-10-01 | 2021-03-19 | 华中农业大学 | Spectral fingerprint identification method for milk, mare, camel, goat and buffalo milk |
| CN112666112A (en) * | 2020-10-01 | 2021-04-16 | 华中农业大学 | Batch discrimination model and method for camel milk and mare milk |
| CN112666114A (en) * | 2020-10-01 | 2021-04-16 | 华中农业大学 | Method for identifying buffalo milk and mare milk by using spectrum |
| CN112666110A (en) * | 2020-10-01 | 2021-04-16 | 华中农业大学 | Spectral identification model and method for milk and goat milk |
| CN112666111A (en) * | 2020-10-01 | 2021-04-16 | 华中农业大学 | Method for quickly identifying milk and mare milk |
| CN112666113A (en) * | 2020-10-01 | 2021-04-16 | 华中农业大学 | Method for quickly identifying goat milk and mare milk |
| CN113049528A (en) * | 2021-02-02 | 2021-06-29 | 浙江理工大学 | Fiber component identification method and module based on near infrared spectrum |
| CN114780917A (en) * | 2022-06-20 | 2022-07-22 | 中国民航大学 | Milk powder doping non-directional detection method based on flow discrimination model |
| CN116165165A (en) * | 2023-04-25 | 2023-05-26 | 四川威斯派克科技有限公司 | Detection method for online real-time release of raw and auxiliary materials of medicines |
| CN116165165B (en) * | 2023-04-25 | 2023-07-07 | 四川威斯派克科技有限公司 | Detection method for online real-time release of raw and auxiliary materials of medicines |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN107024450A (en) | A kind of method for differentiating different brands and hop count milk powder based on near-infrared spectrum technique | |
| Teye et al. | Innovative and rapid analysis for rice authenticity using hand-held NIR spectrometry and chemometrics | |
| Manley et al. | Spectroscopic technique: Near infrared (NIR) spectroscopy | |
| Shen et al. | Discrimination between Shaoxing wines and other Chinese rice wines by near-infrared spectroscopy and chemometrics | |
| Wu et al. | Classification of apple varieties using near infrared reflectance spectroscopy and fuzzy discriminant c‐means clustering model | |
| Shao et al. | Determination of the bruise degree for cherry using Vis-NIR reflection spectroscopy coupled with multivariate analysis | |
| CN102967578A (en) | Method for obtaining near-infrared spectrum of beef sample online and application thereof in evaluating beef quality | |
| CN107219184A (en) | A kind of meat discrimination method and device traced to the source applied to the place of production | |
| Chen et al. | New approach of simultaneous, multi‐perspective imaging for quantitative assessment of the compactness of grape bunches | |
| CN106053383A (en) | Near-infrared online detection method for tobacco processing process | |
| Sharma et al. | Rapid ripening stage classification and dry matter prediction of durian pulp using a pushbroom near infrared hyperspectral imaging system | |
| CN113310936A (en) | Rapid identification method for four high-temperature sterilized commercial milks | |
| Wang et al. | Development of near‐infrared online grading device for long jujube | |
| Jiang et al. | Rapid nondestructive detecting of wheat varieties and mixing ratio by combining hyperspectral imaging and ensemble learning | |
| Wu et al. | Identification of tea varieties by mid‐infrared diffuse reflectance spectroscopy coupled with a possibilistic fuzzy c‐means clustering with a fuzzy covariance matrix | |
| CN106338488A (en) | Method for fast undamaged determination of transgenic soybean milk powder | |
| CN113310937A (en) | Method for rapidly identifying high-temperature sterilized milk, pasteurized fresh milk of dairy cow and reconstituted milk of milk powder | |
| CN113310929A (en) | Soybean powder doped in high-temperature sterilized milk and spectral identification method of doping proportion thereof | |
| CN106485049B (en) | A kind of detection method of the NIRS exceptional sample based on Monte Carlo cross validation | |
| Anzanello et al. | Wavelength selection framework for classifying food and pharmaceutical samples into multiple classes | |
| CN106940292A (en) | Bar denier wood raw material quick nondestructive discrimination method of damaging by worms based on multi-optical spectrum imaging technology | |
| Xu et al. | Detection of apple varieties by near‐infrared reflectance spectroscopy coupled with SPSO‐PFCM | |
| CN113310933A (en) | Spectrum identification method for number of days for storing raw buffalo milk | |
| Hongyang et al. | Detection of browning of fresh‐cut potato chips based on machine vision and electronic nose | |
| CN113310928A (en) | Method for rapidly identifying high-temperature sterilized milk with shelf life within and out of date |
Legal Events
| Date | Code | Title | Description |
|---|---|---|---|
| PB01 | Publication | ||
| PB01 | Publication | ||
| SE01 | Entry into force of request for substantive examination | ||
| SE01 | Entry into force of request for substantive examination | ||
| RJ01 | Rejection of invention patent application after publication |
Application publication date: 20170808 |
|
| RJ01 | Rejection of invention patent application after publication |