CN104330384A - Method for detecting amino acid content in grain by use of terahertz frequency-domain spectrum technology - Google Patents

Abstract

The invention provides a method for detecting amino acid content in grain by use of terahertz frequency-domain spectrum technology. The method comprises the steps of grinding grain samples to be tested and then tableting the grain samples to be tested, directly testing the grain samples to be tested by use of a terahertz time-domain spectrum system in a transmission measurement mode under a nitrogen atmosphere to obtain the terahertz time-domain spectrum signals of the samples, performing Fourier transform on the time-domain spectrums to obtain the frequency-domain spectrums of the samples, performing normalization spectrum preprocessing on all the obtained frequency-domain spectrums, and finally, dividing the preprocessed frequency-domain spectrums of the grain tableted samples to be tested into calibration set frequency-domain spectrums and verification set frequency-domain spectrums, and establishing a quantitative analysis model by use of the partial least squares regression method so as to obtain the quantitative detection values of the various grain samples to be tested. The method for detecting the amino acid content in the grain by use of the terahertz frequency-domain spectrum technology is simple in sample preparation, simple to operate, and capable of truly and effectively realizing fast and accurate quantitative detection on the amino acids in the grain.

Description

The method of application Terahertz frequency domain spectra technology for detection Amino-Acid in Grain content

Technical field

The present invention relates to a kind of method applying Terahertz frequency domain spectra technology for detection Amino-Acid in Grain content, belong to food composition detection technique field.

Background technology

Amino acid is the basic composition unit of polypeptide and biological function macro-molecular protein in constituting body, gives protein specific molecular morphosis, makes protein molecule have biochemical activity.Therefore amino acid has indispensable status in vital movement, is also the necessary important nutrient of growth in humans.Much amino acid not only has unique physiological function, also plays the effect of its uniqueness in the food industry simultaneously.Applying more amino acid in the food industry has glutamic acid, lysine, threonine, tryptophane, arginine etc. at present.Wherein, Pidolidone is due to the flavor of its uniqueness, and mainly for the production of monosodium glutamate, spices etc., and Pidolidone is also important nutritional supplement and biochemical reagents, itself also as medicine, can participate in brain internal protein and carbohydate metabolism, accelerating oxidation process.Therefore, need to find and a kind of detect amino acid whose detection method in food fast and accurately, thus provide foundation for the evaluation of amino acid content in food.

Chinese patent literature CN102590129A discloses a kind of method detecting amino acid content in peanut, concrete steps are as follows: (1) Standard for Peanuts product to known amino acid content carry out near infrared spectrum scanning, obtain all spectral informations of Standard for Peanuts product at near-infrared wavelength of described known amino acid content, obtain the calculating mean value of calibration set sample spectrum; (2) pre-service is carried out to described step (1) gained calibration set sample spectrum; (3) principal component analysis (PCA) is carried out, characteristic information extraction data to the information data in described step (2) pretreated calibration set sample spectrum; (4) with the chemical measurements of the amino acid content of described Standard for Peanuts product for corrected value, using described step (3) gained characteristic information data as independent variable, described corrected value, as dependent variable, sets up the calibration model between described independent variable and described dependent variable with Chemical Measurement Multivariate Correction algorithm; (5) the Standard for Peanuts product of described step (1) described known amino acid content are replaced with peanut sample to be measured, repeating said steps (1)-step (3), described step (3) gained characteristic information data is inputted the calibration model of described step (4), obtain the amino acid content in described peanut sample to be measured.But, above-mentioned employing near infrared spectrum detects the method for amino acid content in peanut, its Problems existing is: near infrared spectrum is that to measure be that frequency multiplication and sum of fundamental frequencies based on intramolecule vibration absorbs, and the spectrum peak therefore presented is wider, and overlap seriously and absorption intensity is weak.Because the spectrum peak of near infrared spectrum is wider, directly can not recognize the individual features absorption peak of material near infrared spectrum middle finger, thus utilize during near-infrared spectrum analysis and need to utilize full modal data Modling model, and then easily introduce more noise information.The penetrability of near infrared spectrum is relatively weak simultaneously, cannot penetrate and there is certain thickness coating or external packing, for when detecting below coating or there is the sample of external packing, often need to destroy coating or external packing, thus destroy the integrality of sample to a certain extent.

Terahertz emission (also claiming " THz radiation ") refers to that frequency is at 0.1THz-10THz, the electromagnetic wave of wavelength between 0.03-3mm, its wave band is between microwave and infrared ray, be the region of macroelectronics to the transition of microcosmic photonics, in electromagnetic spectrum, occupy very special position.The transition of the large molecule of much polarity between vibrational energy level is just in time within Terahertz frequency range, therefore, the tera-hertz spectra of biomolecule, include frequency domain spectra and absorption spectrum, can reflect by molecule or the intrinsic property of low frequency diaphragm that causes of intermolecular collective vibration and lattice vibration.Terahertz electromagnetic wave has lower photon energy, when carrying out sample detection, can not produce harmful photoionization.In addition, THz wave has stronger penetrability, and can penetrate certain thickness interlayer and detect analysis sample, can avoid destroying precious sample, as artifact etc., be a kind of novel effective lossless detection method.But, also the correlation technique of tera-hertz spectra is not used for the relevant report that amino acid detects at present.

Summary of the invention

Technical matters to be solved by this invention is that providing a kind of applies the method that Terahertz frequency domain spectra technology realizes directly carrying out Amino-Acid in Grain quantitatively detection.

For solving the problems of the technologies described above, the present invention is achieved by the following technical solutions:

The method of application Terahertz frequency domain spectra technology for detection Amino-Acid in Grain content, comprises the steps:

(1) get the rear compressing tablet of grain samples to be measured grinding, obtain grain press sheet compression to be measured;

(2) utilize terahertz time-domain spectroscopy system, sample time domain signal, as sample time domain signal, can be obtained the frequency-region signal of sample, i.e. frequency domain spectra by the terahertz time-domain spectroscopy gathering described grain press sheet compression to be measured after Fourier transform.And select the interval validity feature wave band as described target amino acid of the good wave band of reappearance in described frequency domain spectra according to the character of target amino acid;

(3) in selected validity feature wave band, the frequency domain spectra of described grain press sheet compression to be measured is divided into calibration set sample frequency domain spectra and checking collection sample frequency domain spectra; ;

(4) utilize partial least-squares regression method to set up the Quantitative Analysis Model of described calibration set sample frequency domain spectra and described checking collection sample frequency domain spectra, obtain amino acid whose quantitative detected value in each described grain samples to be measured.

Described amino acid is Pidolidone.

Described grain samples is rice.

In described step (2), described validity feature wave band is 0.1-2.0THz.

In described step (2), the test condition of described terahertz time-domain spectroscopy system is: temperature is 20 DEG C, and frequency range is 0.1-3.0THz.

In described step (3), before the frequency domain spectra of described grain press sheet compression to be measured being carried out calibration set and the division of checking collection, also comprise the step of described frequency domain spectra being carried out to Pretreated spectra, described preprocessing procedures is normalization conversion, described normalization transformation range is between 0-1, and concrete transform is as follows:

$x_i^{\′}=\frac{x_i-x_{\min }}{x_{\max }-x_{\min }}$

Here x ' _ibe through the frequency domain amplitude of the rear frequency i of normalization conversion, x _ifor the frequency domain amplitude of original frequency domain spectrum medium frequency i; x _minand x _maxunder frequency i respectively, the minimum frequency domain amplitude in the original frequency domain modal data of all samples and maximum frequency domain amplitude.

In described step (3), the division of described calibration set sample frequency domain spectra and described checking collection sample frequency domain spectra adopts Latin partition method to carry out.

Select partition number to be 4 when utilizing described Latin partition method to divide, get wherein 3/4 as calibration set sample, 1/4 as checking collection sample.

The circle sheet that described grain press sheet compression to be measured is diameter 13mm, thickness is about 1.0mm.

In described step (4), adopt the main cause subnumber of cross-validation method determination partial least squares regression.

In described step (4), adopt square error MSE and square coefficient R of test set ²evaluate the estimated performance accuracy of institute's Modling model, described MSE and R ²computing formula as follows

$\mathrm{MSE}=\frac{1}{n}{\mathrm{\Σ}}_{i=1}^n{(f\left(x_i\right)-y_i)}^2$

$R^2=\frac{{(n{\mathrm{\Σ}}_{i=1}^{n}f\left(x_i\right)y_i-{\mathrm{\Σ}}_{i=1}^{n}f\left(x_i\right){\mathrm{\Σ}}_{i=1}^n{y}_i)}^2}{(n{\mathrm{\Σ}}_{i=1}^{n}f{\left(x_i\right)}^2-{\left({\mathrm{\Σ}}_{i=1}^{n}f\left(x_i\right)\right)}^2)(n{\mathrm{\Σ}}_{i=1}^n{y}_i^2-{\left({\mathrm{\Σ}}_{i=1}^n{y}_i\right)}^2)}$

Technique scheme of the present invention has the following advantages compared to existing technology:

(1) method of application Terahertz frequency domain spectra technology for detection Amino-Acid in Grain content of the present invention, by the grain press sheet compression to be measured that compressing tablet after grain samples grinding to be measured is obtained, direct employing terahertz time-domain spectroscopy system is in transmission measurement pattern, under nitrogen atmosphere, it is tested, obtain the terahertz time-domain spectroscopy signal of sample as sample time domain signal, time-domain signal is after Fourier transform, obtain the frequency domain spectra of described grain press sheet compression to be measured, direct afterwards good for reappearance in frequency domain spectra wave band is decided to be validity feature wave band, finally the frequency domain spectra in the validity feature wave band of described grain press sheet compression to be measured is divided into calibration set sample frequency domain spectra and checking collection sample frequency domain spectra, Quantitative Analysis Model is set up by partial least squares regression, obtain the quantitative detected value of each described grain samples to be measured, the method of the invention can directly utilize the effective information of frequency domain spectra signal to carry out quantitative test, without the need to being further converted to absorption coefficient spectrum, thus simplify data processing step, and do not need to mix other any materials in the sample that detects, sample preparation is simple, do not need to carry out any pre-service, can truly, effectively realize quantitatively detecting fast and accurately the amino acid in grain or food, the prediction square error of described Quantitative Analysis Model is less, squared correlation coefficient (R ²) up to 0.9873.

In addition, the inventive method adopts Terahertz frequency domain spectra technology, and obtains frequency domain spectra after directly Time Domain Spectrum being carried out Fourier transform, and its process does not need the thickness measuring press sheet compression, thus issuable error when avoiding introducing detect thickness.

(2) method of application Terahertz frequency domain spectra technology for detection Amino-Acid in Grain content of the present invention, utilizing before partial least square method sets up calibration model, Latin partition method is adopted to become calibration set frequency domain spectra and checking collection frequency domain spectra to described frequency domain spectra Data Placement, thus can accurate evaluation quantitative calibration models predictive ability and stability; Latin partition method described here is a kind of model performance verification method be based upon on cross validation and random sampling checking basis.Divide process in, described Latin partition method can realize uniform random sampling checking, every partition once, each sample as and only as one-time authentication collection; When partition number is 4, each sample be used for and only for one-time authentication collection, and three times as calibration set, set up four models respectively, ensure that each sample is concentrated at calibration set and checking to occur with same ratio, thus realize evaluating without inclined institute's established model predictive ability, make quantitative calibration models more reliable, analysis result has more statistical significance.

(3) method of application Terahertz frequency domain spectra technology for detection Amino-Acid in Grain content of the present invention, in described step (4), to described calibration set sample frequency domain spectra and described checking collection sample frequency domain spectra, and corresponding concentration value, partial least squares regression is adopted to set up Quantitative Analysis Model, and adopt leaving-one method as the main cause subnumber of cross validation method determination partial least squares regression, thus sample nearly all in every bout is all for training pattern; The most close with the distribution of original sample; Analysis result accuracy is high; Owing to there is not any enchancement factor affecting experimental data in experimentation, experimentation can be replicated; And under the condition of best main cause subnumber, set up Partial Least-Squares Regression Model, farthest can retain the useful information in original spectral data, introduce measurement noises as few as possible simultaneously.

Accompanying drawing explanation

In order to make content of the present invention be more likely to be clearly understood, below according to a particular embodiment of the invention and by reference to the accompanying drawings, the present invention is further detailed explanation, wherein

Fig. 1 is the Terahertz frequency domain spectra of Pidolidone-polyethylene mixture compressing tablet within the scope of 0.1-3.0THz measured in embodiment;

Fig. 2 is the original frequency domain spectrum of part of the present invention rice sample to be measured in 0.1-2.0THz characteristic wave bands interval;

Fig. 3 is the spectrogram of frequency domain spectra described in Fig. 2 after normalization conversion process.

Fig. 4 be described in the root-mean-square error of leave one cross validation and the graph of a relation of PLS main cause subnumber;

Fig. 5 is the graph of a relation between the model predication value of Pidolidone in rice sample to be measured of the present invention and experiment value.

Embodiment

Embodiment

The present embodiment provides a kind of method applying Pidolidone content in Terahertz frequency domain spectra technology for detection grain.Wherein, the grain samples to be measured of Different L-aminoglutaric acid concentration is prepared after rice powder (Zhangjiakou City inspection and quarantine bureau of Hebei province provides) interpolation known quantity Pidolidone (being purchased from Sigma company), and adopt the inventive method to carry out detection analysis as blind sample described grain samples to be measured, concrete steps are as follows:

(1) described rice is put into comminutor to pulverize, sieve after grinding (100 orders, particle diameter≤150 μm), put into baking oven and dry, and obtain dry rice powder, rice powder is mixed according to different quality ratio with Pidolidone powder, be transferred in agate mortar to grind further and obtain described biased sample fine powder to be measured, wherein, in described biased sample to be measured, Pidolidone mass percentage is followed successively by 0.00%, 0.50%, 1.00%, 1.50%, 2.00%, 2.50%, 2.96%, 3.50%, 4.00%, 4.27%, 5.00%, 5.50%, 6.00%, 6.45%, 7.01%, 7.50%, 8.01%, 8.50%, 9.02%, 9.51%, 10.00%, 11.01%, 12.02%, 13.07%, 14.04%, 14.88%, 16.00%, 17.02%, 18.00%, 19.02%, 20.04%, amount to 31 groups,

Take above-mentioned often kind of described grain samples fine powder to be measured and be about 170mg, be placed in the mould of Specac company, under the pressure of 5t, 3-4min is kept with sheeter, thus by the described circle sheet that grain samples fine powder to be measured is pressed into diameter 13mm, thickness is about 1.0mm, obtain the grain press sheet compression to be measured containing different quality number percent Pidolidone, described grain press sheet compression two to be measured surface be parallel, smooth surface and do not have crack;

(2) the terahertz time-domain spectroscopy system of standard is utilized, T-Spectroscopy signals collecting software, under employing U.S. Newport house flag, the Mai Tai titanium sapphire femtosecond laser oscillator of Spectra-Physics brand is as lasing light emitter, pulse center wavelength 800nm, output power > 500mW, under 20 DEG C of conditions, the survey frequency scope of terahertz time-domain spectroscopy system is 0.1-3.0THz, with nitrogen as a reference, adopt transmission measurement pattern, gather the terahertz time-domain spectroscopy signal of described grain press sheet compression to be measured as sample time domain signal, sample time domain signal obtains the frequency-region signal of sample after Fourier transform,

When carrying out above-mentioned data acquisition, the sample spectra of each described grain press sheet compression to be measured repeated acquisition 3 times under equivalent environment, the mean value finally getting 3 times is as subsequent treatment Time Domain Spectrum data used, and final conversion obtains 31 groups of frequency domain spectra;

In order to clearly illustrate the frequency domain amplitude characteristic of Pidolidone in terahertz wave band, adopt polyethylene powders (a kind of thinning agent that THz wave is not almost absorbed and bonding agent, be purchased from Sigma company) mix by the ratio uniform of about 10:1 as medium and pure Pidolidone after, according to the condition tabletted of step (1), measure the terahertz time-domain signal of this compressing tablet afterwards under the same conditions, and convert corresponding frequency-region signal to.

Be illustrated in figure 1 the Terahertz frequency domain spectra of the Pidolidone-polyethylene mixture compressing tablet obtained after Fourier transform, can find out, in the frequency range of 0.1-3.0THz, there is obvious frequency domain amplitude variations in frequency 1.23THz place, can be used as the Terahertz frequency domain character of Pidolidone for qualitative and quantitative analysis.

Be illustrated in figure 2 the original frequency domain spectrum of the grain samples to be measured of 8 Different L-glutamic acid mass percentage among described grain samples to be measured, these 8 Pidolidone mass percentage are respectively 0.00%, 1.00%, 5.00%, 7.01%, 10.00%, 12.01%, 16.00%, 20.04%; Due in the wave band of 0.1-2.0THz, in described frequency domain spectra, reappearance is better, comprises again the frequency domain spectra information of the overwhelming majority simultaneously, and therefore the wave band of 0.1-2.0THz is as the validity feature wave band of described target amino acid; In addition, as can be seen from Figure 2, in the scope that Pidolidone massfraction is 0%-20%, be positioned at frequency domain character and the change in direct ratio of its massfraction of the Pidolidone at 1.23THz place, thus the Terahertz characteristic frequency domain spectrum detecting Pidolidone can realize the quantitative test of Pidolidone in grain.

(3) in the validity feature wave band of selected grain press sheet compression described to be measured, be normalized conversion spectrum pre-service to the frequency domain spectra of all samples, normalized scope is between 0-1, and it is as follows that it specifically counts transform:

$x_i^{\′}=\frac{x_i-x_{\min }}{x_{\max }-x_{\min }}$

Here x ' _ibe through the frequency domain amplitude of the rear frequency i of normalization conversion, x _ifor the original frequency domain amplitude of frequency i.X _minand x _maxunder frequency i respectively, the minimum frequency domain amplitude in all samples original frequency domain data and maximum frequency domain amplitude.

What Fig. 3 showed is through the frequency domain spectrogram that normalization converts rear section sample.Represented concentration is respectively: 0.00%, 1.00%, 5.00%, 7.01%, 10.00%, 12.01%, 16.00%, 20.04%.Similarly, Fig. 3 can indicate more significantly in the scope that Pidolidone massfraction is 0%-20%, the frequency domain character of Pidolidone and the change in direct ratio of its massfraction, so the Terahertz characteristic frequency domain spectrum of pretreated Pidolidone also can be used in the quantitative test of Pidolidone in grain.

(4) in the validity feature wave band of selected grain press sheet compression described to be measured, by the frequency domain spectra after the normalized of described grain press sheet compression to be measured, i.e. above-mentioned 31 groups of pretreated frequency domain spectra, self-service Latin partition method is adopted to be divided into calibration set sample frequency domain spectra and checking collection sample frequency domain spectra, partition number is selected to be 4, get wherein 3/4 as calibration set sample, 1/4 as checking collection sample, be specially: first sample to be tested is divided into 4 parts, select wherein 1 part of conduct checking collection sample, all the other 3 parts as calibration set sample, it should be noted that, in each calculating, each sample is only for once predicting checking.

(5) calibration set is utilized, quantitative calibration models is set up in conjunction with partial least squares regression, adopt leave one cross validation, determine the main cause subnumber of partial least squares regression, as shown in Figure 4, when main gene is 3, the root-mean-square error of cross validation is minimum, RMSECV=0.0063 (MSECV=3.969 × 10 ^-5), adopt 3 main genes when therefore setting up quantitative calibration models.The concrete principle of described partial least squares regression is as follows:

First partial least squares regression decomposes the Terahertz frequency domain spectra matrix X of sample and concentration matrix Y, and its model tormulation is as follows:

$Y={\mathrm{UQ}}^T+E_Y=\underset{i=1}{\overset{k}{\mathrm{\Σ}}}{u}_k{q}_K^T+E_Y$

$X={\mathrm{TP}}^T+E_X=\underset{i=1}{\overset{k}{\mathrm{\Σ}}}{t}_k{p}_k^T+E_X$

In above-mentioned expression formula, t _k(n × 1) is the score of i-th main gene of frequency domain spectra matrix X; p _k(1 × m) is the load of i-th main gene of frequency domain spectra matrix X; u _k(n × 1) is the score of i-th main gene of concentration matrix Y, q _k(1 × p) is the load of i-th main gene of concentration matrix Y; K is main cause subnumber.T and U is the score matrix of X and Y matrix respectively, P and Q is the loading matrix of X and Y matrix respectively, Ex and E _ythen the PLS regression criterion matrix of X and Y matrix respectively.

Afterwards T and U two score matrix T and U are done linear regression:

U＝TB

B＝(T ^TT) ^-1T ^TY

Last when predicting, first obtain the score matrix T ＇ of unknown sample frequency domain spectra matrix X ＇ according to the loading matrix P of frequency domain spectra matrix X, then can be tried to achieve the concentration prediction matrix Y ＇ of unknown sample by following formula:

Y＇＝T＇BQ

The concentration prediction value y ＇ of the unknown sample in concentration prediction matrix Y ＇ can be obtained thus;

(6) utilize checking collection frequency domain spectra, verify the estimated performance of offset minimum binary quantitative calibration models set up, the actual concentrations of different sample and the comparing result of prediction concentrations as shown in table 1; Result shows, and within the scope of the Pidolidone massfraction of 0-20%, the prediction square error of partial least square model is MSE=4.2133 × 10 ^-5, squared correlation coefficient R ²=0.9872.

Be illustrated in figure 5 described model about the graph of a relation between the predicted value of Pidolidone and experiment value, can know and find out that the correlativity between described model predication value and experiment value is satisfactory.Squared correlation coefficient reaches 0.9872.

The different sample actual concentrations value of table 1-and prediction concentrations value

As can be seen here, above-mentioned analytical model prediction square error (MSEP) and predicted root mean square error (RMSEP) less, R ²up to 0.9873, thus illustrate that the model that the inventive method is set up is reliably feasible, can be used for quantitatively detecting Pidolidone in rice sample.

Adopt the method for foregoing description, not only the Pidolidone in rice quantitatively can be detected, also can be generalized to the amino acid whose detection of other kind.In the detection of other kind grain or food samples, also need sample preparation powdered compressing tablet.As long as and the target amino acid being suitable for detecting has characteristic absorption or changing features in Terahertz frequency range and available the method for the invention realizes detecting.Detection method of the present invention can be accurate, easy the aminoacid ingredient in grain is detected, obtained the subsidy of the great scientific instrument special project (2012YQ140005) of state natural sciences fund (21275101) and country.

Obviously, above-described embodiment is only for clearly example being described, and the restriction not to embodiment.For those of ordinary skill in the field, can also make other changes in different forms on the basis of the above description.Here exhaustive without the need to also giving all embodiments.And thus the apparent change of extending out or variation be still among the protection domain of the invention.

Claims (10)

1. apply the method for Terahertz frequency domain spectra technology for detection Amino-Acid in Grain content, it is characterized in that, comprise the steps:

(1) get the rear compressing tablet of grain samples to be measured grinding, obtain grain press sheet compression to be measured;

(2) terahertz time-domain spectroscopy system is applied, adopt the terahertz time-domain spectroscopy signal of the grain press sheet compression to be measured under its nitrogen atmosphere of transmission measurement type collection, sample time domain spectrum directly obtains the frequency domain spectra of sample after Fourier transform, and selects the interval validity feature wave band as described target amino acid of the good wave band of reappearance in described frequency domain spectra according to the character of target amino acid:

(3) in selected characteristic wave bands, the frequency domain spectra of described grain press sheet compression to be measured is divided into calibration set sample frequency domain spectra and checking collection sample frequency domain spectra;

(4) utilize partial least-squares regression method to set up the Quantitative Analysis Model of described calibration set sample frequency domain spectra and described checking collection sample frequency domain spectra, obtain amino acid whose quantitative detected value in each described grain samples to be measured.

2. the method for application Terahertz frequency domain spectra technology for detection Amino-Acid in Grain content according to claim 1, it is characterized in that, described amino acid is Pidolidone.

3. the method for application Terahertz frequency domain spectra technology for detection Amino-Acid in Grain content according to claim 2, it is characterized in that, described grain samples is rice.

4. eat the method for amino acid content in application Terahertz frequency domain spectra technology for detection grain according to claim 3, it is characterized in that, in described step (2), the validity feature wave band of described target amino acid is 0.1-2.0THz.

5. according to the method for the arbitrary described application Terahertz frequency domain spectra technology for detection Amino-Acid in Grain content of claim 1-4, it is characterized in that, in described step (2), the test condition of described terahertz time-domain spectroscopy system is: temperature is 20 DEG C, and frequency range is 0.1-3.0THz.

6. according to the method for the arbitrary described application Terahertz frequency domain spectra technology for detection grain amino acid content of claim 1-5, it is characterized in that, in described step (3), before the frequency domain spectra of described grain press sheet compression to be measured being divided training set and checking collection, also comprise the step described frequency domain spectra being done to Pretreated spectra, the method of described Pretreated spectra is normalization conversion, and described normalization transformation range is between 0-1, and concrete transform is as follows:

$x_i^{\′}=\frac{x_i-x_{\min }}{x_{\max }-x_{\min }}$

Here x ' _ibe through the frequency domain amplitude of the rear frequency i of normalization conversion, x _ifor the frequency domain amplitude of original frequency domain spectrum medium frequency i.X _minand x _maxunder frequency i respectively, the minimum frequency domain amplitude in all samples original frequency domain data and maximum frequency domain amplitude.

7. according to the method for the arbitrary described application Terahertz frequency domain spectra technology for detection Amino-Acid in Grain content of claim 1-6, it is characterized in that, in described step (3), the division of described calibration set sample frequency domain spectra and described checking collection sample frequency domain spectra adopts Latin partition method to carry out.

8. according to the method for the arbitrary described application Terahertz frequency domain spectra technology for detection Amino-Acid in Grain content of claim 1-7, it is characterized in that, select partition number to be 4 when utilizing described Latin partition method to divide, get wherein 3/4 as calibration set sample, 1/4 as checking collection sample.

9. according to the method for the arbitrary described application Terahertz frequency domain spectra technology for detection Amino-Acid in Grain content of claim 1-8, it is characterized in that, in described step (4), adopt the main cause subnumber that cross validation determination partial least square method returns.

10. according to the method for the arbitrary described application Terahertz frequency domain spectra technology for detection Amino-Acid in Grain content of claim 1-9, it is characterized in that, in described step (5), adopt square error MSE and square coefficient R of checking collection ²evaluate the estimated performance accuracy of institute's Modling model, described MSE and R ²computing formula as (1) and (2):

$\mathrm{MSE}=\frac{1}{n}{\mathrm{\Σ}}_{i=1}^n{(f\left(x_i\right)-y_i)}^2---\left(1\right)$

$R^2=\frac{{(n{\mathrm{\Σ}}_{i=1}^{n}f\left(x_i\right)y_i-{\mathrm{\Σ}}_{i=1}^{n}f\left(x_i\right){\mathrm{\Σ}}_{i=1}^n{y}_i)}^2}{(n{\mathrm{\Σ}}_{i=1}^{n}f{\left(x_i\right)}^2-{\left({\mathrm{\Σ}}_{i=1}^{n}f\left(x_i\right)\right)}^2)(n{\mathrm{\Σ}}_{i=1}^n{y}_i^2-{\left({\mathrm{\Σ}}_{i=1}^n{y}_i\right)}^2)}---\left(2\right)$