CN106706552A - Method for detecting wheat quality based on multiple-information integration - Google Patents
Method for detecting wheat quality based on multiple-information integration Download PDFInfo
- Publication number
- CN106706552A CN106706552A CN201611059135.0A CN201611059135A CN106706552A CN 106706552 A CN106706552 A CN 106706552A CN 201611059135 A CN201611059135 A CN 201611059135A CN 106706552 A CN106706552 A CN 106706552A
- Authority
- CN
- China
- Prior art keywords
- wheat
- sample
- samples
- refractive index
- adaboost
- 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
- 241000209140 Triticum Species 0.000 title claims abstract description 179
- 235000021307 Triticum Nutrition 0.000 title claims abstract description 179
- 238000000034 method Methods 0.000 title abstract description 14
- 230000010354 integration Effects 0.000 title abstract 2
- 238000001228 spectrum Methods 0.000 claims abstract description 53
- 238000000862 absorption spectrum Methods 0.000 claims abstract description 45
- 230000004927 fusion Effects 0.000 claims description 33
- 238000012360 testing method Methods 0.000 claims description 33
- 238000012549 training Methods 0.000 claims description 27
- 238000001514 detection method Methods 0.000 claims description 22
- 230000003595 spectral effect Effects 0.000 claims description 18
- 238000012706 support-vector machine Methods 0.000 claims description 17
- 238000004458 analytical method Methods 0.000 claims description 16
- 238000005516 engineering process Methods 0.000 claims description 11
- 238000013145 classification model Methods 0.000 claims description 10
- 238000010521 absorption reaction Methods 0.000 claims description 9
- 238000000605 extraction Methods 0.000 claims description 7
- 235000013399 edible fruits Nutrition 0.000 claims description 3
- 238000010606 normalization Methods 0.000 claims description 3
- 230000009466 transformation Effects 0.000 claims description 3
- 230000008901 benefit Effects 0.000 abstract description 5
- 230000003287 optical effect Effects 0.000 abstract description 4
- 238000000513 principal component analysis Methods 0.000 abstract description 3
- 238000012163 sequencing technique Methods 0.000 abstract 1
- 239000000463 material Substances 0.000 description 6
- 238000005259 measurement Methods 0.000 description 3
- 238000013507 mapping Methods 0.000 description 2
- 241000196324 Embryophyta Species 0.000 description 1
- 238000013528 artificial neural network Methods 0.000 description 1
- 230000009286 beneficial effect Effects 0.000 description 1
- 239000003153 chemical reaction reagent Substances 0.000 description 1
- 230000000052 comparative effect Effects 0.000 description 1
- 230000007123 defense Effects 0.000 description 1
- 238000010586 diagram Methods 0.000 description 1
- 230000005611 electricity Effects 0.000 description 1
- 230000003628 erosive effect Effects 0.000 description 1
- 238000002474 experimental method Methods 0.000 description 1
- 238000002329 infrared spectrum Methods 0.000 description 1
- 238000007689 inspection Methods 0.000 description 1
- 239000000155 melt Substances 0.000 description 1
- 238000012986 modification Methods 0.000 description 1
- 230000004048 modification Effects 0.000 description 1
- 238000007500 overflow downdraw method Methods 0.000 description 1
- 238000012797 qualification Methods 0.000 description 1
- 238000003908 quality control method Methods 0.000 description 1
- 238000011002 quantification Methods 0.000 description 1
- 238000004445 quantitative analysis Methods 0.000 description 1
- 238000011160 research Methods 0.000 description 1
- 239000000126 substance Substances 0.000 description 1
- 239000002699 waste material 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
- G01N21/3581—Investigating relative effect of material at wavelengths characteristic of specific elements or molecules, e.g. atomic absorption spectrometry using infrared light using far infrared light; using Terahertz radiation
- G01N21/3586—Investigating relative effect of material at wavelengths characteristic of specific elements or molecules, e.g. atomic absorption spectrometry using infrared light using far infrared light; using Terahertz radiation by Terahertz time domain spectroscopy [THz-TDS]
-
- 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
Abstract
The invention discloses a method for detecting wheat quality based on multiple-information integration. The method comprises the following steps of 1, using a terahertz time-domain spectrum technique THz-TDS to analyze the optical and spectrum characteristics of mildewing, worm-eaten, sprouted and normal wheat samples at the wave band of 0.2 to 1.6THz, and calculating to obtain the absorption spectrum and refractive index spectrum of the wheat samples; 2, using a PCA (principal component analysis) method to extract the features of the absorption spectrum and refractive index spectrum of the wheat samples, and sequencing the feature values according to size; 3, selecting the previous eight principal component features to be combined as a wheat sample absorption spectrum feature set, and selecting the previous ten principal component features to be combined as a wheat quality refractive index spectrum feature set; 4, modeling, and verifying the modeling result. The method has the advantages that the absorption spectrum and refractive index spectrum information of the wheat samples with different qualities are integrated, and the identifying rate of the wheat samples is improved.
Description
Technical field
The present invention relates to a kind of detection method of wheat quality.It is more particularly related to a kind of be based on multi information
The wheat quality detection method of fusion.
Background technology
The report to the detection method of wheat quality of storing in a warehouse mainly has both at home and abroad：Chemical method (measurement main component), electricity
The methods such as sub- nose, machine vision, spectral detection (infrared spectrum), these methods waste time and energy, big to sample and reagent consumption, inspection
Survey limited in one's ability.THz (Terahertz), because its photon energy is low and composes unique advantages such as " fingerprints ", is a kind of effective point
The non-cpntact measurement means of analysis material inside composition information.THz-TDS is current most representational THz technologies, in biology
The fields such as medical science, material science, national defense safety and quality control have important application.
THz ripples in terms of the quantification and qualification of material, many scholars around material THz wave bands optics and light
Spectrum signature, has carried out the analysis work of the THz absorption coefficients, the measurement of specific refractivity and single optical parameter of material.Existing skill
Art combination THz spectrum and chemometrics method (principal component analysis, offset minimum binary, SVMs etc.) are realized to material
Quantitative analysis, while also yielding good result.Other researcher obtains different wheat samples using THz-TDS technologies
THz optical parametrics, are recognized using PCR, PLS, BP neural network and PCA-SVM models to sample, use wheat samples
THz absorption spectrums set up wheat quality identification model be respectively 50% to the discrimination of different quality wheats, 58.33%,
83.33% and 93.33%.The discrimination of normal wheat and germinated wheat is relatively high, and is mostly in the sample of erroneous judgement
Mouldy wheat and worm-eaten wheat, it is larger that different models carry out Classification and Identification rate difference.
The content of the invention
It is an object of the invention to solve at least the above, and provide the advantage that at least will be described later.
It is a still further object of the present invention to provide a kind of wheat quality detection method based on Multi-information acquisition, it can be tied
The Multi-source Information Fusion model of absorption spectrum and refractive index spectra joint mapping is closed, the discrimination of wheat samples is substantially increased,
Its overall discrimination has reached more than 95%.
In order to realize these purposes of the invention and further advantage, there is provided a kind of wheat based on Multi-information acquisition
Quality detecting method, it is comprised the following steps：
Step one, using terahertz time-domain spectroscopic technology THz-TDS, analysis is gone mouldy, damages by worms, germinateing and normal wheat
Sample 0.2~1.6THz wave bands optics and spectral characteristic, and calculate obtain wheat samples absorption spectrum and refractive index light
Spectrum；
Different quality wheat samples absorption spectrums and refractive index spectra are entered by step 2 using PCA principal component analytical methods
Row feature extraction, characteristic value is ranked up according to size；
Step 3, it is wheat samples absorption spectrum feature set to select preceding 8 principal component combinations of features, first 10 of selection it is main into
Dtex is levied and is combined as wheat quality refractive index spectra feature set；
Step 4, modeling, and modeling result is verified.
Preferably, being modeled as in the step 5 sets up the wheat fusion for classification model based on AdaBoost, specifically
Step is as follows：
1) the wheat samples spectral information being calculated the step one is random according to 2：1 is divided into training set and test
Collection；
2) 4 graders of AdaBoost wheats two are set up：By the training set according to 1：1 ratio is randomly divided into calibration set
And forecast set, set up normal wheat grader, germinated wheat grader, mouldy wheat grader and worm-eaten respectively using calibration set
4 graders of AdaBoost bis- of wheat grader；Wherein, the iterations for setting the graders of AdaBoost bis- is 100-200 times；
3) output result according to 4 graders of AdaBoost bis- carries out judgement classification to wheat samples, if output result
Then will be positive sample more than or equal to 0, otherwise be then negative sample, and output result absolute value is bigger, and degree of belief is with regard to highest.
4) combine go mouldy, damage by worms, germinateing and normal wheat samples THz absorption spectrums and refractive index spectra, set up
Based on AdaBoost wheats categorised decision layer Fusion Model；Mould is merged based on AdaBoost wheats categorised decision layer using setting up
Type is identified to the sample in test set, and the test data input graders of AdaBoost bis- are differentiated and calculated point
The accuracy of class result.
Preferably, being modeled as in the step 5 sets up SVM wheat fusion for classification models, comprises the following steps that：
(1) training sample and test sample collection of wheat samples are set up：The wheat samples that the step one is calculated
Absorption spectrum and refractive index spectra it is random according to 2：1 is divided into training set and test set；
(2) it is wheat samples absorption spectrum feature set to select preceding 8 principal component combinations of features, selects preceding 10 principal component spies
Levy and be combined as wheat quality refractive index spectra feature set, be modeled using RBF kernel functions, by specific refractivity and absorption coefficient
SVMs is built respectively as RBF kernel functions to be trained, complete the modeling of SVMs；
(3) sample in test set is identified using the SVM models set up, test data is input into SVMs
Differentiated and calculated the accuracy of classification results.
Preferably, from the kernel function that RBF kernel functions are Fusion Model in the step (2), while kernel function is optimal
The selection of parameter γ, c is calculated by grid search optimized algorithm and realized.
Preferably, the RBF kernel functions, are expressed as：
Need to be by repetition test, analytical error adjusts kernel functional parameter C, γ, searches out optimal value；Use root-mean-square error
(RMSE) precision of prediction of final regression model is assessed.RMSE is represented by：
Wherein, N " represents the quantity of training set sample；YiWithIth samples actual value and ith in data set are represented respectively
Sample predicted value in the regression model for building.
Preferably, the step 3) in wheat samples are carried out according to the output result of 4 graders of AdaBoost bis-
Judge that classification is concretely comprised the following steps：
A, if A1,A2,A3,A4Respectively 4 outputs of the graders of AdaBoost bis-；
B, for different quality wheat sample Xi, wherein 1 ＜ i ＜ N, N represents the quantity of wheat sample, according to what is set up
The graders of AdaBoost bis- are differentiated, obtain 4 graders outputs, are expressed as A1(Xi),A2(Xi),A3(Xi),A4
(Xi)；
C, compares Aj(Xi) value, obtain output valve it is maximum when grader jmax=MAX { Aj(Xi), 1≤j≤4 }, then
Can be by XiIt is jmaxClass sample；
D, travels through all samples, obtains the classification results of all samples.
Preferably, the graders of Adaboost bis- are achieved by the steps of：
A, it is assumed that X represents sample space, Y is sample class logo collection, then sample training collection is S={ (xi,yi| i=1,
2,…,m)},xi∈X,yi∈Y
B, initializes the m weights of training sample, and initial sample weights are distributed as：D1(i)=1/m,
C, in the probability distribution D of given training samplet(i)Under, train Weak Classifier ht；
Calculate the weighting fault rate of Weak ClassifierDt(i)Represent imparting in the t times iteration
The weights of sample, t represents iterations；
More new formula is
The weights distribution of more new training sample set, for next iteration
Wherein ZtIt is normalization factor,
D, strong classifier prediction output result.
Preferably, in the step one, optics and spectral characteristic of the wheat samples in 0.2~1.6THz wave bands are obtained
Afterwards, the frequency domain spectra of sample is obtained using Fourier transformation, and calculates acquisition THz absorption coefficients and refractive index, set up small
The absorption spectrum and refractive index spectra of wheat sample.
Preferably, the wheat quality detection method based on Multi-information acquisition is comprised the following steps：
Step one, using terahertz time-domain spectroscopic technology THz-TDS, analysis is gone mouldy, damages by worms, germinateing and normal wheat
Sample and calculates the spectral information for obtaining wheat samples in the optics and spectral characteristic of 0.2~1.6THz wave bands, spectrum letter
Breath includes absorption spectrum and refractive index spectra；
Different quality wheat samples absorption spectrums and refractive index spectra are entered by step 2 using PCA principal component analytical methods
Row feature extraction, characteristic value is ranked up according to size；
Step 3, it is wheat samples absorption spectrum feature set to select preceding 8 principal component combinations of features, first 10 of selection it is main into
Dtex is levied and is combined as wheat quality refractive index spectra feature set；
Step 4, is modeled using RBF kernel functions, using specific refractivity and absorption coefficient as RBF kernel functions
Build SVMs to be trained, complete the modeling of SVMs；Using the SVM models set up to the sample in test set
It is identified, test data input SVMs is differentiated and calculated the accuracy of classification results.
The present invention at least includes following beneficial effect：Wheat quality detection side based on Multi-information acquisition of the present invention
Method, absorption spectrum and refractive index spectra from wheat breed carry out the structure of classification and Detection Fusion Model.SVM is established respectively
Wheat fusion for classification model and AdaBoost Fusion Models are identified to wheat samples.Carried out from multinomial spectral target first
Modeling, substantially increases the discrimination of wheat samples, and its overall discrimination has reached more than 95%.Wheat breed of the present invention
The optical parametric such as absorption spectrum and refractive index spectra utilizes THz-TDS technologies, selects the optics and light of 0.2~1.6THz wave bands
Spectral property is used as modeling basis.Substantially increase the discrimination of SVM wheat fusion for classification models and AdaBoost Fusion Models.
Further advantage of the invention, target and feature embody part by following explanation, and part will also be by this
The research and practice of invention and be understood by the person skilled in the art.
Brief description of the drawings
Fig. 1 is the curve map of AdaBoost graders iterations correspondence forecast set error in classification of the present invention；
Fig. 2 is AdaBoost, SVM Fusion Model of the present invention and the wheat Classification and Identification knot using PCA-SVM models
Fruit comparison diagram.
Specific embodiment
The present invention is described in further detail below in conjunction with the accompanying drawings, to make those skilled in the art with reference to specification text
Word can be implemented according to this.
It should be appreciated that it is used herein such as " have ", "comprising" and " including " term is not precluded from one or many
The presence or addition of individual other elements or its combination.
Wheat quality detection method based on Multi-information acquisition of the present invention is comprised the following steps：
Step one, using terahertz time-domain spectroscopic technology THz-TDS, analysis is gone mouldy, damages by worms, germinateing and normal wheat
Sample 0.2~1.6THz wave bands optics and spectral characteristic, and calculate obtain wheat samples absorption spectrum and refractive index light
Spectrum；
Different quality wheat samples absorption spectrums and refractive index spectra are entered by step 2 using PCA principal component analytical methods
Row feature extraction, characteristic value is ranked up according to size；
Step 3, it is wheat samples absorption spectrum feature set to select preceding 8 principal component combinations of features, first 10 of selection it is main into
Dtex is levied and is combined as wheat quality refractive index spectra feature set；
Step 4, modeling, and modeling result is verified.
Embodiment 1
Wheat quality detection method based on Multi-information acquisition of the present invention is comprised the following steps：
S101, using terahertz time-domain spectroscopic technology THz-TDS, analysis is gone mouldy, damages by worms, germinateing and normal wheat sample
Product 0.2~1.6THz wave bands optics and spectral characteristic, and calculate obtain wheat samples absorption spectrum and refractive index spectra；
Different quality wheat samples absorption spectrums and refractive index spectra are carried out by S102 using PCA principal component analytical methods
Feature extraction, characteristic value is ranked up according to size；
S103, it is wheat samples absorption spectrum feature set to select preceding 8 principal component combinations of features, selects preceding 10 principal components
Combinations of features is wheat quality refractive index spectra feature set；
S104, modeling is random according to 2 by the absorption spectrum and refractive index spectra of wheat samples：1 is divided into training set and test
Collection；Set up 4 graders of AdaBoost wheats two：By the training set according to 1：1 ratio is randomly divided into calibration set and prediction
Collection, normal wheat grader, germinated wheat grader, mouldy wheat grader and worm-eaten wheat point are set up using calibration set respectively
4 graders of AdaBoost bis- of class device；Wherein, the iterations for setting the graders of AdaBoost bis- is 200 times；Adaboost bis-
Grader is achieved by the steps of：
A, it is assumed that X represents sample space, Y is sample class logo collection, then sample training collection is
S={ (xi,yi| i=1,2 ..., m) }, xi∈X,yi∈Y
B, initializes the m weights of training sample, and initial sample weights are distributed as：D1(i)=1/m,
C, in the probability distribution D of given training samplet(i)Under, train Weak Classifier ht；
Calculate the weighting fault rate of Weak ClassifierDt(i)Represent imparting in the t times iteration
The weights of sample, t represents iterations；
More new formula is
The weights distribution of more new training sample set, for next iteration
Wherein ZtIt is normalization factor,
D, strong classifier prediction output result.
S105, the output result according to 4 graders of AdaBoost bis- carries out judgement classification to wheat samples, if output knot
Fruit is more than or equal to 0, then will be positive sample, otherwise is then negative sample, and output result absolute value is bigger, and degree of belief is with regard to highest.
Judge that sorting technique is specific as follows：
1) A is set1,A2,A3,A4Respectively 4 outputs of the graders of AdaBoost bis-；
2) for different quality wheat sample Xi, wherein 1 ＜ i ＜ N, N represents the quantity of wheat sample, according to what is set up
The graders of AdaBoost bis- are differentiated, obtain 4 graders outputs, are expressed as A1(Xi),A2(Xi),A3(Xi),A4
(Xi)；
3) A is comparedj(Xi) value, obtain output valve it is maximum when grader jmax=MAX { Aj(Xi), 1≤j≤4 }, then
Can be by XiIt is jmaxClass sample；
4) all samples are traveled through, the classification results of all samples are obtained.
S106, with reference to go mouldy, damage by worms, germinate and normal wheat samples THz absorption spectrums and refractive index spectra, build
Be based on AdaBoost wheats categorised decision layer Fusion Model；Merged based on AdaBoost wheats categorised decision layer using setting up
Model is identified to the sample in test set, and the test data input graders of AdaBoost bis- are differentiated and calculated
The accuracy of classification results.
The different iterations of selection, records four error changes of the graders of AdaBoost bis-, as a result such as Fig. 1.By Fig. 1
Can obtain, with the increase of iterations, error is gradually reduced.When iterations is less than 100 times, fluctuate also obvious, error
It is relatively large；At 100-150 times, the error of 4 graders tends towards stability iterations substantially, wherein, take second place more than 150
Afterwards, error line is consistent；When iterations is too big, the increase of model complexity will cause the estimated performance to reduce.Therefore originally
Invention chooses iteration 200 times.And the error of normal wheat grader and germinated wheat grader than worm-eaten wheat class device and
The error of mouldy wheat grader is low.
Wheat samples are identified using AdaBoost wheats categorised decision layer Fusion Model in test set, wheat samples
Overall discrimination is 95%, and table 1 gives the discrimination and erroneous judgement quantity of various wheat samples.
14 kinds of AdaBoost wheat fusion for classification model modeling results of table
By Biao Ke get, the knowledge of the AdaBoost Fusion Models of foundation to the normal wheat samples of test set and germinated wheat sample
Not rate is stronger, and respectively 100%, the equally discrimination to mouldy wheat samples and worm-eaten wheat samples is relatively low, respectively
90.48% and 91.3%.
Embodiment 2
Wheat quality detection method based on Multi-information acquisition of the present invention is comprised the following steps：
S201, using terahertz time-domain spectroscopic technology THz-TDS, analysis is gone mouldy, damages by worms, germinateing and normal wheat sample
Product 0.2~1.6THz wave bands optics and spectral characteristic, and calculate obtain wheat samples absorption spectrum and refractive index spectra；
Such as Fourier transformation obtains the frequency domain spectra of sample, and calculates acquisition THz absorption coefficients and refractive index.
Different quality wheat samples absorption spectrums and refractive index spectra are carried out by S202 using PCA principal component analytical methods
Feature extraction, characteristic value is ranked up according to size；
S203, it is wheat samples absorption spectrum feature set to select preceding 8 principal component combinations of features, selects preceding 10 principal components
Combinations of features is wheat quality refractive index spectra feature set；
S204, sets up SVM wheat fusion for classification models, comprises the following steps that：
(1) training sample and test sample collection of wheat samples are set up：The wheat samples that the step one is calculated
Absorption spectrum and refractive index spectra it is random according to 2：1 is divided into training set and test set；
(2) it is wheat samples absorption spectrum feature set to select preceding 8 principal component combinations of features, selects preceding 10 principal component spies
Levy and be combined as wheat quality refractive index spectra feature set, be modeled using RBF kernel functions, by specific refractivity and absorption coefficient
SVMs is built respectively as RBF kernel functions to be trained, complete the modeling of SVMs；
(3) sample in test set is identified using the SVM models set up, test data is input into SVMs
Differentiated and calculated the accuracy of classification results.
From the kernel function that RBF kernel functions are Fusion Model in the step (2), while kernel function optimized parameter γ, c's
Selection is calculated by grid search optimized algorithm and realized.The RBF kernel functions, are expressed as：
Need to be by repetition test, analytical error adjusts kernel functional parameter C, γ, searches out optimal value；Use root-mean-square error
(RMSE) precision of prediction of final regression model is assessed.RMSE is represented by：
Wherein, N " represents the quantity of training set sample；YiWithIth samples actual value and ith in data set are represented respectively
Sample predicted value in the regression model for building.
Test data input SVMs in test set is differentiated and calculated the accuracy of classification results.
The results are shown in Table 2
The SVM wheat fusion for classification model modeling results of table 2
By Biao Ke get, the SVM Fusion Models of foundation are to the normal wheat samples of test set, germinated wheat sample, mouldy wheat
The discrimination of sample and worm-eaten wheat samples is respectively 100%, 100%, 95.24% and 96.65%.Result shows that SVM is merged
Model is all higher to the discrimination of 4 kinds of wheat samples, is a kind of more satisfactory wheat fusion for classification model.
Comparative example 1
Using the PCA-SVM models set up according to absorption spectrum merely to normal wheat, germinated wheat, mouldy wheat, worm
Erosion wheat is identified.
To the absorption spectrum and refractive index light of 4 kinds of samples such as normal wheat, germinated wheat, mouldy wheat and worm-eaten wheat
Spectrum, is utilized respectively the classification and Detection Fusion Model of the multinomial spectral target of wheat that AdaBoost graders and SVM models are set up, two
Plant the discrimination of model quite, and the precision of prediction of the PCA-SVM models set up than single absorption spectrum and refractive index spectra has
Certain raising.Additionally, Fusion Model can accurately identify normal wheat and germinated wheat sample, SVM sets up and melts
Matched moulds type is slightly above AdaBoost Fusion Models to mouldy wheat and worm-eaten wheat samples discrimination.
As seen from Figure 2, the PCA-SVM models set up according to absorption spectrum merely are to normal wheat, germinated wheat, mouldy
Wheat, worm-eaten wheat and overall discrimination are respectively 100%, 100%, 85.71%, 82.61% and 92.08%；Simple basis
The PCA-SVM models that refractive index spectra is set up are respectively to the discrimination and overall discrimination of this wheat samples in 4：100%th,
100%th, 80.95%, 86.96% and 91.98%, and the overall discrimination of Fusion Model has reached more than 95%.Experiment table
It is bright, for different quality wheat samples, with reference to absorption spectrum and refractive index spectra joint mapping Multi-source Information Fusion one
Determine to be improve in degree the discrimination of wheat samples, and SVM Feature-level fusions discrimination highest, SVM Feature-level fusions are
Optimal multi-sources Information Fusion Method.
Although embodiment of the present invention is disclosed as above, it is not restricted to listed in specification and implementation method
With, it can be applied to various suitable the field of the invention completely, for those skilled in the art, can be easily
Other modification is realized, therefore under the universal limited without departing substantially from claim and equivalency range, the present invention is not limited
In specific details and shown here as the legend with description.
Claims (9)
1. a kind of wheat quality detection method based on Multi-information acquisition, it is characterised in that comprise the following steps：
Step one, using terahertz time-domain spectroscopic technology THz-TDS, analysis is gone mouldy, damages by worms, germinateing and normal wheat samples
In the optics and spectral characteristic of 0.2~1.6THz wave bands, and calculate the absorption spectrum and refractive index spectra for obtaining wheat samples；
Different quality wheat samples absorption spectrums and refractive index spectra are carried out spy by step 2 using PCA principal component analytical methods
Extraction is levied, characteristic value is ranked up according to size；
Step 3, it is wheat samples absorption spectrum feature set to select preceding 8 principal component combinations of features, selects preceding 10 principal component spies
Levy and be combined as wheat quality refractive index spectra feature set；
Step 4, modeling, and modeling result is verified.
2. the wheat quality detection method of Multi-information acquisition is based on as claimed in claim 1, it is characterised in that the step 4
In be modeled as set up based on AdaBoost wheat fusion for classification model, comprise the following steps that：
1) the wheat samples spectral information being calculated the step one is random according to 2：1 is divided into training set and test set；
2) 4 graders of AdaBoost wheats two are set up：By the training set according to 1：1 ratio is randomly divided into calibration set and pre-
Collection is surveyed, normal wheat grader, germinated wheat grader, mouldy wheat grader and worm-eaten wheat is set up respectively using calibration set
4 graders of AdaBoost bis- of grader；Wherein, the iterations for setting the graders of AdaBoost bis- is 100-200 times；
3) output result according to 4 graders of AdaBoost bis- carries out judgement classification to wheat samples, if output result is more than
Then will be positive sample equal to 0, otherwise be then negative sample, and output result absolute value is bigger, and degree of belief is with regard to highest.
4) combine go mouldy, damage by worms, germinateing and normal wheat samples THz absorption spectrums and refractive index spectra, foundation is based on
AdaBoost wheats categorised decision layer Fusion Model；Using foundation based on AdaBoost wheats categorised decision layer Fusion Model pair
Sample in test set is identified, and the test data input graders of AdaBoost bis- are differentiated and classification knot is calculated
The accuracy of fruit.
3. the wheat quality detection method of Multi-information acquisition is based on as claimed in claim 1, it is characterised in that the step 4
In be modeled as set up SVM wheat fusion for classification models, comprise the following steps that：
(1) training sample and test sample collection of wheat samples are set up：The suction of the wheat samples that the step one is calculated
Receive spectrum and refractive index spectra is random according to 2：1 is divided into training set and test set；
(2) it is wheat samples absorption spectrum feature set to select preceding 8 principal component combinations of features, selects preceding 10 principal component feature groups
Wheat quality refractive index spectra feature set is combined into, is modeled using RBF kernel functions, specific refractivity and absorption coefficient are distinguished
SVMs is built as RBF kernel functions to be trained, complete the modeling of SVMs；
(3) sample in test set is identified using the SVM models set up, test data input SVMs is carried out
Differentiate and calculate the accuracy of classification results.
4. the wheat quality detection method of Multi-information acquisition is based on as claimed in claim 3, it is characterised in that the step
(2) from the kernel function that RBF kernel functions are Fusion Model in, while the selection of kernel function optimized parameter γ, c passes through grid search
Optimized algorithm is calculated to be realized.
5. the wheat quality detection method of Multi-information acquisition is based on as claimed in claim 4, it is characterised in that the RBF cores
Function, is expressed as：
Need to be by repetition test, analytical error adjusts kernel functional parameter C, γ, searches out optimal value；Use root-mean-square error
(RMSE) precision of prediction of final regression model is assessed.RMSE is represented by：
Wherein, N " represents the quantity of training set sample；YiWithIth samples actual value and ith samples in data set are represented respectively
The predicted value in the regression model for building.
6. the wheat quality detection method of Multi-information acquisition is based on as claimed in claim 2, it is characterised in that the step 3)
It is middle wheat samples to be carried out according to 4 output results of the graders of AdaBoost bis- judge that classification is concretely comprised the following steps：
A, if A1,A2,A3,A4Respectively 4 outputs of the graders of AdaBoost bis-；
B, for different quality wheat sample Xi, wherein 1 ＜ i ＜ N, N represents the quantity of wheat sample, according to what is set up
The graders of AdaBoost bis- are differentiated, obtain 4 graders outputs, are expressed as A1(Xi),A2(Xi),A3(Xi),A4
(Xi)；
C, compares Aj(Xi) value, obtain output valve it is maximum when grader jmax=MAX { Aj(Xi), 1≤j≤4 }, then can be by
XiIt is jmaxClass sample；
D, travels through all samples, obtains the classification results of all samples.
7. the wheat quality detection method of Multi-information acquisition is based on as claimed in claim 6, it is characterised in that Adaboost bis-
Grader is achieved by the steps of：
A, it is assumed that X represents sample space, Y is sample class logo collection, then sample training collection is S={ (xi,yi| i=1,
2,…,m)},xi∈X,yi∈Y
B, initializes the m weights of training sample, and initial sample weights are distributed as：D1(i)=1/m,
C, in the probability distribution D of given training samplet(i)Under, train Weak Classifier ht；
Calculate the weighting fault rate of Weak ClassifierDt(i)Represent and sample is assigned in the t times iteration
Weights, t represents iterations；
More new formula is
The weights distribution of more new training sample set, for next iteration
Wherein ZtIt is normalization factor,
D, strong classifier prediction output result.
8. the wheat quality detection method of Multi-information acquisition is based on as claimed in claim 1, it is characterised in that the step one
In, wheat samples are obtained after the optics and spectral characteristic of 0.2~1.6THz wave bands, sample is obtained using Fourier transformation
Frequency domain spectra, and calculate acquisition THz absorption coefficients and refractive index, set up the absorption spectrum and refractive index spectra of wheat samples.
9. the wheat quality detection method of Multi-information acquisition is based on as claimed in claim 1, it is characterised in that including following step
Suddenly：
Step one, using terahertz time-domain spectroscopic technology THz-TDS, analysis is gone mouldy, damages by worms, germinateing and normal wheat samples
In the optics and spectral characteristic of 0.2~1.6THz wave bands, and calculate the spectral information for obtaining wheat samples, the spectral information bag
Include absorption spectrum and refractive index spectra；
Different quality wheat samples absorption spectrums and refractive index spectra are carried out spy by step 2 using PCA principal component analytical methods
Extraction is levied, characteristic value is ranked up according to size；
Step 3, it is wheat samples absorption spectrum feature set to select preceding 8 principal component combinations of features, selects preceding 10 principal component spies
Levy and be combined as wheat quality refractive index spectra feature set；
Step 4, is modeled using RBF kernel functions, and specific refractivity and absorption coefficient are built as RBF kernel functions
SVMs is trained, and completes the modeling of SVMs；The sample in test set is carried out using the SVM models set up
Identification, test data input SVMs is differentiated and is calculated the accuracy of classification results.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201611059135.0A CN106706552A (en) | 2016-11-25 | 2016-11-25 | Method for detecting wheat quality based on multiple-information integration |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201611059135.0A CN106706552A (en) | 2016-11-25 | 2016-11-25 | Method for detecting wheat quality based on multiple-information integration |
Publications (1)
| Publication Number | Publication Date |
|---|---|
| CN106706552A true CN106706552A (en) | 2017-05-24 |
Family
ID=58934803
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201611059135.0A Pending CN106706552A (en) | 2016-11-25 | 2016-11-25 | Method for detecting wheat quality based on multiple-information integration |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN106706552A (en) |
Cited By (5)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN107884428A (en) * | 2017-10-31 | 2018-04-06 | 重庆楚新业科技有限公司 | A kind of dangerous article detection system and its control method |
| CN107991265A (en) * | 2017-11-17 | 2018-05-04 | 西南大学 | A kind of wheat flour Rubus biflorus Buch quick determination method based on information fusion |
| CN108458989A (en) * | 2018-04-28 | 2018-08-28 | 江苏建筑职业技术学院 | A kind of Coal-rock identification method based on Terahertz multi-parameter spectrum |
| CN110163101A (en) * | 2019-04-17 | 2019-08-23 | 湖南省中医药研究院 | The difference of Chinese medicine seed and grade quick discrimination method |
| CN110458362A (en) * | 2019-08-15 | 2019-11-15 | 中储粮成都储藏研究院有限公司 | Grain quality index prediction technique based on SVM supporting vector machine model |
-
2016
- 2016-11-25 CN CN201611059135.0A patent/CN106706552A/en active Pending
Non-Patent Citations (2)
| Title |
|---|
| 葛宏义: "储粮品质的THz波检测理论与分析方法研究", 《中国科学院大学博士学位论文》 * |
| 赵源深等: "番茄采摘机器人非颜色编码化目标识别算法研究", 《农业机械学报》 * |
Cited By (7)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN107884428A (en) * | 2017-10-31 | 2018-04-06 | 重庆楚新业科技有限公司 | A kind of dangerous article detection system and its control method |
| CN107991265A (en) * | 2017-11-17 | 2018-05-04 | 西南大学 | A kind of wheat flour Rubus biflorus Buch quick determination method based on information fusion |
| CN108458989A (en) * | 2018-04-28 | 2018-08-28 | 江苏建筑职业技术学院 | A kind of Coal-rock identification method based on Terahertz multi-parameter spectrum |
| CN108458989B (en) * | 2018-04-28 | 2020-10-09 | 江苏建筑职业技术学院 | Terahertz multi-parameter spectrum-based coal rock identification method |
| CN110163101A (en) * | 2019-04-17 | 2019-08-23 | 湖南省中医药研究院 | The difference of Chinese medicine seed and grade quick discrimination method |
| CN110163101B (en) * | 2019-04-17 | 2022-09-23 | 湖南省中医药研究院 | Method for rapidly distinguishing seeds of traditional Chinese medicinal materials and rapidly judging grades of seeds |
| CN110458362A (en) * | 2019-08-15 | 2019-11-15 | 中储粮成都储藏研究院有限公司 | Grain quality index prediction technique based on SVM supporting vector machine model |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN106706552A (en) | Method for detecting wheat quality based on multiple-information integration | |
| CN102313699B (en) | Estimation method of total nitrogen content in crop canopy leaf | |
| Brown et al. | The origin and global distribution of second order variability in satellite ocean color and its potential applications to algorithm development | |
| CN104020135B (en) | Calibration model modeling method based near infrared spectrum | |
| Liu et al. | A method combining ELM and PLSR (ELM-P) for estimating chlorophyll content in rice with feature bands extracted by an improved ant colony optimization algorithm | |
| Mishra et al. | A Deep Learning-Based Novel Approach for Weed Growth Estimation. | |
| Mishra et al. | Identification of citrus greening (HLB) using a VIS-NIR spectroscopy technique | |
| CN104535528B (en) | The method of BP neural network extract real-time TDLAS gas absorption spectra absorbances | |
| CN104298999B (en) | EO-1 hyperion feature learning method based on recurrence autocoding | |
| CN104392251A (en) | Hyperspectral image classification method based on semi-supervised dictionary learning | |
| CN103983584A (en) | Retrieval method and retrieval device of chlorophyll a concentration of inland case II water | |
| Nurmasari et al. | Oil palm plantation detection in Indonesia using Sentinel-2 and Landsat-8 optical satellite imagery (case study: Rokan Hulu regency, Riau Province) | |
| CN102982339A (en) | Hyperspectral characteristic variable selection method | |
| Wang et al. | Winter wheat chlorophyll content retrieval based on machine learning using in situ hyperspectral data | |
| CN103278467A (en) | Rapid nondestructive high-accuracy method with for identifying abundance degree of nitrogen element in plant leaf | |
| US20050111692A1 (en) | Cloud cover assessment: VNIR-SWIR | |
| CN103456011A (en) | Improved hyperspectral RX abnormal detection method by utilization of complementary information | |
| CN103743679B (en) | A kind of method accelerating greenhouse gases vertical column retrieving concentration speed | |
| CN103955711B (en) | A kind of mode identification method in imaging spectral target identification analysis | |
| CN104359847B (en) | Method and device for acquiring centroid set used for representing typical water category | |
| Ma et al. | Applications of a hyperspectral imaging system used to estimate wheat grain protein: a review | |
| CN106778815A (en) | Wheat quality THz spectral classification methods based on DS evidence theories | |
| Zhou et al. | Estimation of volume resources for planted forests using an advanced LiDAR and hyperspectral remote sensing | |
| Liu et al. | Diagnosis of citrus greening using raman spectroscopy-based pattern recognition | |
| CN108663334A (en) | The method for finding soil nutrient spectral signature wavelength based on multiple Classifiers Combination |
Legal Events
| Date | Code | Title | Description |
|---|---|---|---|
| PB01 | Publication | ||
| PB01 | Publication | ||
| 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: 20170524 |