CN104297202B - The method for quantitatively detecting Pesticide Residues In Grain using THz TDS frequency domain spectras - Google Patents
The method for quantitatively detecting Pesticide Residues In Grain using THz TDS frequency domain spectras Download PDFInfo
- Publication number
- CN104297202B CN104297202B CN201410510382.2A CN201410510382A CN104297202B CN 104297202 B CN104297202 B CN 104297202B CN 201410510382 A CN201410510382 A CN 201410510382A CN 104297202 B CN104297202 B CN 104297202B
- Authority
- CN
- China
- Prior art keywords
- sample
- frequency domain
- grain
- domain spectra
- training set
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Active
Links
Abstract
The present invention relates to a kind of method that utilization THz TDS frequency domain spectras quantitatively detect Pesticide Residues In Grain, tabletting obtains grain press sheet compression to be measured after by the way that grain samples to be measured are ground, it is tested using THz TDS, obtain terahertz time-domain spectroscopy, it is fourier transformed and obtains frequency domain spectra, the selected preferable wave band interval of frequency domain spectra reappearance is used as characteristic wave bands, the grain press sheet compression to be measured is training set sample frequency domain spectra in the frequency domain spectra random division of characteristic wave bands and collection sample frequency domain spectra is verified, the Quantitative Analysis Model of frequency domain spectra is set up with partial least-square regression method, obtain the quantitative detected value of each grain samples to be measured, the method of the invention can be true, effectively realize and Pesticide Residues In Grain is fast and accurately quantitatively detected, the average value of the prediction related coefficient of the frequency domain spectra Quantitative Analysis Model is up to 0.995.
Description
Technical field
The present invention relates to a kind of method that utilization THz-TDS frequency domain spectras quantitatively detect Pesticide Residues In Grain, belong to agricultural chemicals
Detection technique field.
Background technology
With the fast development of modern agriculture, agricultural and agricultural product are growing to the demand of agricultural chemicals and dependence.In addition one
A little practitioners lack pesticide knowldgde, long-term largely even to be abused using agricultural chemicals so that agricultural chemicals is caused to environment and human health
Strong influence and harm.For crops, often had after the applications of pesticide to crops, in crops a small amount of residual
Stay, remains of pesticide of ingesting for a long time can have a strong impact on health, can also cause significant damage to body and ecological environment.Especially
It is in recent years that teratogenesis caused by food Pesticide Residues are exceeded is disabled and poisoning etc. also increasingly attracts attention, agricultural chemicals
Vestige is remained and its metabolin is found in soil, water and agricultural product.Thus, before scientifical use agricultural chemicals is vigorously advocated
Put, how fast and accurately to detect and just seem particularly important and urgent the problems such as the pollution of food, environment caused by agricultural chemicals
Cut.
There are corresponding method of detection and standard, existing agriculture in countries in the world to agricultural product and Pesticide maximum residue limit
Medicine detection method mainly has gas chromatography, high performance liquid chromatography, internal standard method for gas chromatography technology, liquid chromatogram and matter
GC-MS, immunization, hexavalent chrome bio-removal etc. are composed, but the above method has sample pre-treatments complexity, and detection time is long,
Testing cost is high, requires higher to testing staff, it is impossible to the deficiency such as on-line checking.
Terahertz emission (also referred to as " THz radiation ") refers to frequency in 0.1THz-10THz, and wavelength is between 0.03-3mm
Electromagnetic wave, its wave band be located between microwave and infrared ray, be macroelectronics to the region of microcosmic photonic propulsion transition, in electromagnetic wave
Occupy very special position in frequency spectrum.Transition of many polarity macromoleculars between vibration level is exactly in Terahertz frequency model
Enclose, therefore, the tera-hertz spectra of biomolecule can reflect as caused by intramolecular or intermolecular collective vibration and lattice vibration
The intrinsic property of low frequency diaphragm.Terahertz electromagnetic wave has relatively low photon energy, when carrying out sample detection, will not produce
Harmful photoionization, is a kind of effective lossless detection method.
Chinese patent literature CN103472032A discloses one kind using terahertz time-domain spectroscopic technology detection hydrochloric acid Fourth Ring
The method of element, it comprises the following steps：(1) quadracycline powder is mixed in varing proportions with high-density polyethylene powder and ground
Mill, is pressed into disk-shaped wafers with tablet press machine, obtains the quadracycline tabletting containing different quality percentage one by one；(2) exist
In 0.1-3.5THz band limits, the quadracycline pressure of different quality percentage is gathered one by one with terahertz time-domain spectroscopy system
The terahertz time-domain spectroscopy of piece；(3) using the time domain waveform of nitrogen as reference signal, with the time domain waveform of quadracycline tabletting
As sample signal, Fourier transformation is carried out respectively, the frequency domain distribution of two kinds of signals is obtained, and tabletting sample is obtained using formula
Absorption coefficient and refractive index；(4) calibration model is set up according to the mass percent of each tabletting sample and its corresponding absorption spectra,
It is corrected the checking and evaluation of model.Although the above method can be realized qualitative to antibiotic quadracycline progress in food
But it is by antibiotic quadracycline and high density polyethylene (HDPE) powder to be detected in Sample Preparation Procedure with quantitative detection
Last mixed pressuring plate sample, using polyethylene be " background " detect antibiotic content, thus this method finally detect be exist
The relative value of the content of antibiotic in polyethylene, this is resulted in differs greatly with actual detection sample, have impact on detection
The authenticity of sample；Meanwhile, new foreign matter polyethylene is introduced, will unavoidably shadow be caused to the degree of accuracy of pattern detection
Ring, while also wasting medicine.In addition, more people are the processing using Terahertz absorption spectra now, so that qualitative or fixed
Amount studies agricultural chemicals, without the useful information more fully utilized obtained by THz-TDS technologies.
The content of the invention
The technical problems to be solved by the invention provide a kind of direct to Pesticide Residues In Grain using THz-TDS frequency domain spectras
The method quantitatively detected.
In order to solve the above technical problems, the present invention is achieved by the following technical solutions：
A kind of method that utilization THz-TDS frequency domain spectras quantitatively detect Pesticide Residues In Grain, it is characterised in that including as follows
Step：
(1) tabletting after grain samples grinding to be measured is taken, grain press sheet compression to be measured is produced；
(2) the target pesticide sample to be detected is taken, according to mass ratio 1:4 ratio is mixed with polyethylene, and is ground
Tabletting afterwards, produces the mixed pressuring plate sample containing target agricultural chemicals and polyethylene；
(3) using THz-TDS spectroscopic systems the grain press sheet compression to be measured and the mixed pressuring plate sample are carried out by
One test, the terahertz time-domain spectroscopy signal obtained under the terahertz time-domain spectroscopy of each sample, collection nitrogen atmosphere is used as reference
Signal, and the terahertz time-domain spectroscopy signal of the grain press sheet compression to be measured and mixed pressuring plate sample is gathered under the same conditions
As sample signal, by reference signal and sample signal respectively through the frequency domain spectrum signal Es referred to after Fourier transformation and
The frequency domain spectrum signal Er of sample：
In formula, ω represents frequency,The phase difference of representative sample signal and reference signal, ρ is that sample signal is believed with reference
Number amplitude mode ratio；
(4) according to the frequency domain spectrum signal of the mixed pressuring plate sample, the selected preferable wave band interval of reappearance is used as feature
Wave band, and be training set sample frequency domain by frequency domain spectra random division of the grain press sheet compression to be measured under the characteristic wave bands
Spectrum and checking collection sample frequency domain spectra；
(5) the training set sample frequency domain spectra and the checking collection sample frequency domain are set up using partial least-square regression method
The Quantitative Analysis Model of spectrum, obtains the quantitative detected value of each grain samples to be measured.
The agricultural chemicals is insecticide.
The insecticide is Acetamiprid, and the Acetamiprid frequency domain spectra characteristic wave bands are 0.1-2.0THz.
The insecticide is sevin, and the frequency domain spectra characteristic wave bands of the sevin are 0.5-2.0THz.
The insecticide is imidacloprid, and the frequency domain spectra characteristic wave bands of the imidacloprid are in 0.1-2.0THz.
In the step (5), set up the training set sample frequency domain spectra using partial least-square regression method and described test
The Quantitative Analysis Model of card collection sample frequency domain spectra, the number of principal components of the Acetamiprid frequency domain spectra Quantitative Analysis Model is 2, the west
The number of principal components of denapon frequency domain spectra Quantitative Analysis Model is 4, and the main cause subnumber of the imidacloprid frequency domain spectra Quantitative Analysis Model is
6。
In the step (3), the test condition of the THz-TDS is：Detection temperature is 19 DEG C, with nitrogen as reference,
The scanning step motor interval of spectrometer scanning system is -1mm-2mm, and step-length is 0.01mm.
In the step (4), the division of the training set sample and the checking collection sample matches somebody with somebody point-score using self-service Latin
Carry out；
Specific calculating process of the self-service Latin with point-score is as follows：
(a) assume that data set has n sample, sample is randomly ordered, extract N number of sample out from data set every time afterwards；
(b) data set is divided into the m groups of equal sizes, m=n/N；
(c) using first group as checking collection, remaining m-1 groups are training set, i.e., with n-N training set Sample Establishing analysis mould
Type, and predict the N number of sample for being extracted and；
(d) using second group as checking collection, remaining m-1 groups are training set, i.e., with n-N training set Sample Establishing analysis mould
Type, and predict the N number of sample for being extracted and；
(e) by that analogy, carry out m times altogether, until all groups are extracted once and carry out being predicted as stopping.
When being divided using the self-service Latin with point-score to the frequency domain spectra, repeat partition and calculate 10 times.
In the step (3), each sample duplicate measurements 3 times, each sample be diameter 13mm, thickness 1.2mm circle it is thin
Piece.
In the step (5), the specific Computing Principle of the partial least-square regression method is as follows：
Deflected secondary air is built upon the model on the basis of independent variable X and dependent variable Y matrixes, by setting up independent variable
Linear regression model (LRM) of the latent variable on the latent variable of dependent variable, and then react the relation between independent variable and dependent variable；
X=TPT+ E=∑s tapa t；
Y=UQT+ F=∑s uaqa t；
In formula：T is X score matrix, and P is X loading matrix, and E is X residual matrix, taFor score vector, paFor phase
The load vectors answered, U is Y score matrix, and Q is Y loading matrix, and F is Y residual matrix, uaFor score vector, qaFor phase
The load vectors answered；
PLS extracts respective latent variable in X and Y respectively, and they are respectively independent variable and dependent variable
Linear combination, while following condition should be met：
(a) two groups of latent variables farthest carry the variation information of independent variable and dependent variable respectively；
(b) covariance therebetween is maximized.
The PLS is a method with iterative method progressively extract component, the profit mutually in iterative calculation
With the information of other side, iteration is constantly according to X, Y remaining information adjustment t each timea、uaThe constituents extraction of the second wheel is carried out, until
Element absolute value in remaining matrix is approximately zero, and algorithm stops.Thus obtained coefficient can preferably reflect X and Y relation.
The above-mentioned technical proposal of the present invention has advantages below compared with prior art：
(1) method that utilization THz-TDS frequency domain spectras of the present invention quantitatively detect Pesticide Residues In Grain, by that will treat
The grain press sheet compression to be measured that tabletting is obtained after grain samples grinding is surveyed, directly it is surveyed using THz-TDS spectroscopic systems
Examination, obtains terahertz time-domain spectroscopy, is fourier transformed and obtains frequency domain spectra, afterwards directly by the preferable ripple of reappearance in frequency domain spectra
Duan Dingwei characteristic wave bands, and be training by frequency domain spectra random division of the grain press sheet compression to be measured under the characteristic wave bands
Collect sample frequency domain spectra and checking collection sample frequency domain spectra, Quantitative Analysis Model is set up with partial least-square regression method, each institute is obtained
State the quantitative detected value of grain samples to be measured；The method of the invention can be carried out directly using the effective information of frequency domain spectrum signal
Quantitative analysis, without being further converted to absorption coefficient spectrum, so that data processing step is simplified, and it is described for examining
Be not required to mix other any materials in the sample of survey, sample preparation is simple, can truly, effectively realize it is residual to grain Pesticides
Stay and fast and accurately quantitatively detected, the average value of the prediction related coefficient (Rv) of the Quantitative Analysis Model is up to 0.995.
(2) method that utilization THz-TDS frequency domain spectras of the present invention quantitatively detect Pesticide Residues In Grain, in the step
Suddenly with point-score divide using Latin in (4) obtaining the training set sample frequency domain spectra and the checking collection sample frequency domain spectra,
Itself reason for this is that：The Latin with point-score be testing model robustness a kind of method, in calculating process, first by sample with
Machine sorts, and extracts N number of sample out from data set every time afterwards, the N number of sample being extracted with remaining sample and prediction, protects
Demonstrate,proving each sample is used for and is only used for 1 prediction, it is ensured that the tabletting sample of different quality containing is concentrated in training set and checking
Occur at the same scale, so as to realize the unbiased evaluation to institute's established model predictive ability, make identification model more reliable, analysis knot
Fruit is more statistically significant.
(3) method that utilization THz-TDS frequency domain spectras of the present invention quantitatively detect Pesticide Residues In Grain, in the step
Suddenly in (5), partial least-square regression method is used to the training set sample frequency domain spectra and the checking collection sample frequency domain spectra
Quantitative Analysis Model is set up, and using correction root-mean-square error (RMSEC), predicted root mean square error (RMSEP), prediction phase relation
Number (Rv) foundation judged as model performance, RMSEC, RMSEP be smaller, RvBigger, model is better.
Brief description of the drawings
In order that present disclosure is more likely to be clearly understood, below in conjunction with the accompanying drawings, further detailed is made to the present invention
Thin explanation, wherein,
Fig. 1 is the frequency domain spectra of Acetamiprid described in the embodiment of the present invention 1-polyethylene mixed pressuring plate sample；
Fig. 2 is the interval frequency domain spectra of the grain samples 0.1-2.0THz characteristic wave bands to be measured of part described in the embodiment of the present invention 1；
Fig. 3 is the prediction mean square error of analysis model described in the embodiment of the present invention 1 and the graph of a relation of number of principal components；
Fig. 4 is the graph of a relation between analysis model predicted value and experiment value described in the embodiment of the present invention 1；
Fig. 5 is the frequency domain spectra of sevin described in the embodiment of the present invention 2-polyethylene mixed pressuring plate sample；
Fig. 6 is the interval frequency domain spectra of the grain samples 0.5-2.0THz characteristic wave bands to be measured of part described in the embodiment of the present invention 2；
Fig. 7 is the graph of a relation of analysis model prediction mean square error and number of principal components described in the embodiment of the present invention 2；
Fig. 8 is the graph of a relation between analysis model predicted value and experiment value described in the embodiment of the present invention 2.
Fig. 9 is the Terahertz frequency domain spectra of imidacloprid described in the embodiment of the present invention 3-polyethylene mixed pressuring plate sample；
Figure 10 is frequency domain of the grain samples to be measured described in the embodiment of the present invention 3 in 0.1-2.0THz characteristic wave bands interval
Spectrum；
Figure 11 is the graph of a relation of cross validation root-mean-square error described in the embodiment of the present invention 3 and PLS main cause subnumbers；
Figure 12 is the graph of a relation between model prediction concentration and actual concentrations described in the embodiment of the present invention 3.
Embodiment
Embodiment 1
The present embodiment provides one kind and quantitatively detects pyridine worm in wheat samples using THz-TDS frequency domain spectra combination Chemical Measurements
The method of amidine content, wherein, " wheat samples " are using the cake wheat flour (Binzhou for itself not containing any residues of pesticides
Tai Yumai industry Co., Ltd) prepare after addition known quantity Acetamiprid (offer of Beijing North Na Chuanlian Bioteknologisk Institut),
And " wheat samples " are detected as blind sample using the inventive method by described in, are comprised the following steps that：
(1) cake is put into wheat flour in pulverizer and crushed, sieved (200 mesh), be put into baking oven and dry, obtain
First wheat powder processed；It wheat powder and Acetamiprid powder processed will just mix, be transferred in agate mortar according to different quality ratio
Further grinding obtains described " wheat samples " fine powder, wherein, Acetamiprid weight/mass percentage composition is successively in " wheat samples "
For 0%, 0.5%, 1.0%, 1.5%, 2.0%, 2.5%, 3.0%, 3.5%, 4.0%, 4.5%, 5.5%, 6.0%,
6.5%, 7.0%, 9.0%, 12.0%, 14.0%, 20.0%, 25.0%, 27.0%；
Weigh in above-mentioned every kind of " wheat samples " fine powder about 200mg, the mould for being placed in Specac companies, use tablet press machine
Keep 3-4min under 5t pressure, thus will described in " wheat samples " fine powder be pressed into diameter 13mm, thickness 1.2mm circle it is thin
Piece, produces the grain press sheet compression to be measured containing different quality percentage Acetamiprid, the surface of grain press sheet compression two to be measured
Parallel, surface is smooth and without crack；
(2) by Acetamiprid powder and polyethylene powders according to mass ratio 1:4 mixing, are transferred in agate mortar and further grind
Mill, the mixed pressuring plate sample is obtained using with the compacting of step (1) identical method；
(3) the THz-TDS spectroscopic systems used are described as follows：
Femtosecond laser oscillator exports laser center wavelength 800nm, pulsewidth 35fs, repetition rate 74MHz.Femtosecond laser leads to
Cross after 1/2 wave plate and devating prism, beam energy is divided into two, one is as pump light, the direct irradiation after optical delay line
THz radiation is produced on to large aperture GaAs photoelectric traverses, 4 piece 90 is utilized°Off axis paraboloidal mirror is collected and calibration THz ripples, finally
Focused it onto transmitted through silicon chip on ZnTe crystal；Secondly as probe, after condenser lens is converged and is reflected through film with
The THz ripples of generation are collinearly irradiated on ZnTe crystal, and THz wave is detected using electro optic sampling method.Voltage signal is by locking
Phase amplifier carries out de-noising and enhanced processing, finally sampled with data acquisition software, handle obtain terahertz pulse when
Domain electric field waveform.
With above-mentioned THz-TDS spectroscopic systems to the grain press sheet compression to be measured and (totally 21, the mixed pressuring plate sample
Tabletting) to be tested one by one, each tabletting duplicate measurements obtains the terahertz time-domain spectroscopy of each sample for 3 times, averages；It is described
THz-TDS test condition is：Temperature is 19 DEG C, with nitrogen as reference, the scanning step motor area of spectrometer scanning system
Between be -1mm-2mm, step-length is 0.01mm, and the frequency range of THz-TDS systems is 0.1-3.0THz；
The terahertz time-domain spectroscopy signal under nitrogen atmosphere is gathered as reference signal, and is gathered under the same conditions described
The terahertz time-domain spectroscopy signal of grain press sheet compression and mixed pressuring plate sample to be measured is as sample signal, by reference signal and sample
Frequency domain spectrum signal Er of the product signal respectively through the frequency domain spectrum signal Es and sample referred to after Fourier transformation：
Wherein ω represents frequency,The phase difference of representative sample and reference signal, ρ is sample and the amplitude mode of reference signal
Ratio；
(4) frequency domain spectra as shown in Figure 1 for the Acetamiprid-polyethylene mixed pressuring plate sample, it can be seen that in 0.1-
In 3.0THz wave bands, Acetamiprid is located at frequency 0.74THz, 0.76THz, 0.83THz, 1.18THz, 1.60THz, 1.68THz
In the presence of 6 characteristic absorptions, therefore, selection 0.1-2.0THz is as characteristic wave bands, the frequency domain spectra in the characteristic wave bands are interval
Reappearance is preferable；
Part grain samples to be measured are illustrated in figure 2 in the interval frequency domain spectra of 0.1-2.0THz characteristic wave bands, pyridine worm
The mass percent of amidine is respectively 0.5%, 6%, 7%, 9%, 12%, 25%；
For the frequency domain spectra of 0.1-2.0THz characteristic wave bands, training set sample frequency domain is divided into point-score using Latin
Spectrum and checking collect sample frequency domain spectra, and N values take 5 here；Then 20 samples are randomly divided into 4 groups, 15 samples as training set,
5 samples collect as checking, each partition, ensure that each sample only collects as 1 checking, and each sample is used as training set
3 times, repeat partition 10 times；
Wherein, the Latin matches somebody with somebody point-score, and specific calculating process is described as follows：
(a) assume that data set has n sample, sample is randomly ordered；
(b) data set is divided into the m groups (m=n/N) of equal sizes；
(c) using first group as checking collection, remaining m-1 groups are training set, i.e., with n-N training set Sample Establishing analysis mould
Type, and predict the N number of sample for being extracted and；
(d) using second group as checking collection, remaining m-1 groups are training set, i.e., with n-N training set Sample Establishing analysis mould
Type, and predict the N number of sample for being extracted and；
(e) by that analogy, carry out m times altogether, until all groups are extracted once and carry out being predicted as stopping；
(5) fixed is set up using partial least-square regression method to the training set sample and the checking collection sample frequency domain spectra
Analysis model is measured, its concrete principle is described as follows：
Deflected secondary air is built upon the model on the basis of independent variable X and dependent variable Y matrixes, by setting up independent variable
Linear regression model (LRM) of the latent variable on the latent variable of dependent variable, and then react the relation between independent variable and dependent variable；
X=TPT+ E=∑s tapa t；
Y=UQT+ F=∑s uaqa t；
In formula：T is X score matrix, and P is X loading matrix, and E is X residual matrix, taFor score vector, paFor phase
The load vectors answered, U is Y score matrix, and Q is Y loading matrix, and F is Y residual matrix, uaFor score vector, qaFor phase
The load vectors answered；
PLS extracts respective latent variable in X and Y respectively, and they are respectively independent variable and dependent variable
Linear combination, while following condition should be met：
(a) two groups of latent variables farthest carry the variation information of independent variable and dependent variable respectively；
(b) covariance therebetween is maximized；
PLS is a method with iterative method progressively extract component.The utilization pair mutually in iterative calculation
The information of side, iteration is constantly according to X, Y remaining information adjustment t each timea、uaThe constituents extraction of the second wheel is carried out, until remnants
Element absolute value in matrix is approximately zero, and algorithm stops.Thus obtained coefficient can preferably reflect X and Y relation.
The graph of a relation of the model prediction mean square error and number of principal components is illustrated in figure 3, when number of principal components is 2, mould
The predicted value of type and experiment value are very close to being illustrated in figure 4 the graph of a relation between the model predication value and experiment value, predict phase
Relation number can reach 0.995.
Institute's established model is verified with point-score using Latin, correction root-mean-square error (RMSEC), prediction root mean square are missed
The foundation that poor (RMSEP), prediction related coefficient (Rv) are judged as model performance；Under different partition number of times, obtained school is calculated
Positive root-mean-square error (RMSEC), predicted root mean square error (RMSEP), prediction related coefficient (Rv) are as shown in table 1.
Table 1- PLSs set up the results of property of analysis model
The data from table 1 are again it can be seen that RMSEC, RMSEP of above-mentioned analysis model are smaller, RvIt is larger, so as to say
The model that bright the inventive method is set up is reliably feasible, can be for quantitatively being detected to Acetamiprid in wheat samples.
Embodiment 2
The present embodiment provides one kind and quantitatively detects that rice sample Chinese and Western is tieed up using THz-TDS frequency domain spectra combination Chemical Measurements
Because of content, wherein, " rice sample " using itself do not contain any residues of pesticides rice (Jilin man good fortune rice industry it is limited
Company produces) prepare after addition known quantity sevin (offers of Beijing North Na Chuanlian Bioteknologisk Institut), and will described in
" rice sample " is detected as blind sample using the inventive method, is comprised the following steps that：
(1) rice is put into pulverizer and crushed, sieved (100 mesh), be put into baking oven and dry, obtain rice meal
End；Rice powder and sevin powder are mixed according to different quality ratio, further grinding is transferred in agate mortar and obtains
" rice sample " fine powder, wherein, sevin weight/mass percentage composition is followed successively by 0%, 0.4% in " rice sample ",
0.5%, 0.6%, 0.7%, 0.8%, 0.9%, 1.0%, 2.0%, 3.0%, 4.0%, 5.0%, 6.0%, 7.0%,
8.0%；
Weigh in above-mentioned every kind of " rice sample " fine powder about 200mg, the mould for being placed in Specac companies, use tablet press machine
Keep 3-4min under 5t pressure, thus will described in " rice sample " fine powder be pressed into diameter 13mm, thickness 1.2mm circle it is thin
Piece, produces the grain press sheet compression to be measured containing different quality percentage sevin, the surface of grain press sheet compression two to be measured
Parallel, surface is smooth and without crack；
(2) by sevin powder and polyethylene powders according to mass ratio 1:4 mixing, are transferred in agate mortar and further grind
Mill, the mixed pressuring plate sample is obtained using with the compacting of step (1) identical method；
(3) the THz-TDS spectroscopic systems used are described as follows：
Femtosecond laser oscillator exports laser center wavelength 800nm, pulsewidth 35fs, repetition rate 74MHz.Femtosecond laser leads to
Cross after 1/2 wave plate and devating prism, beam energy is divided into two, one is as pump light, the direct irradiation after optical delay line
THz radiation is produced on to large aperture GaAs photoelectric traverses, 4 piece 90 is utilized°Off axis paraboloidal mirror is collected and calibration THz ripples, finally
Focused it onto transmitted through silicon chip on ZnTe crystal；Secondly as probe, after condenser lens is converged and is reflected through film with
The THz ripples of generation are collinearly irradiated on ZnTe crystal, and THz wave is detected using electro optic sampling method.Voltage signal is by locking
Phase amplifier carries out de-noising and enhanced processing, finally sampled with data acquisition software, handle obtain terahertz pulse when
Domain electric field waveform.
With above-mentioned THz-TDS spectroscopic systems to the grain press sheet compression to be measured and (totally 16, the mixed pressuring plate sample
Tabletting) to be tested one by one, each tabletting duplicate measurements obtains the terahertz time-domain spectroscopy of each sample for 3 times, averages；It is described
THz-TDS test condition is：Temperature is 19 DEG C, with nitrogen as reference, the scanning step motor area of spectrometer scanning system
Between be -1mm-2mm, step-length is 0.01mm, and the frequency range of THz-TDS systems is 0.1-3.0THz；
The terahertz time-domain spectroscopy signal under nitrogen atmosphere is gathered as reference signal, and is gathered under the same conditions described
The terahertz time-domain spectroscopy signal of grain press sheet compression and mixed pressuring plate sample to be measured is as sample signal, by reference signal and sample
Frequency domain spectrum signal Er of the product signal respectively through the frequency domain spectrum signal Es and sample referred to after Fourier transformation：
Wherein ω represents frequency,The phase difference of representative sample and reference signal, ρ is sample and the amplitude mode of reference signal
Ratio；
(4) it is illustrated in figure 5 the frequency domain spectra of the sevin-polyethylene mixed pressuring plate sample, it can be seen that in 0.1-
In 3.0THz wave bands, sevin, which is located at frequency 0.9THz, has 1 characteristic absorption, therefore, and selection 0.5-2.0THz is used as spy
Wave band is levied, the reappearance of frequency domain spectra is preferable in the characteristic wave bands are interval；
It is illustrated in figure 6 the interval frequency domain spectra of part grain samples 0.5-2.0THz characteristic wave bands to be measured, Acetamiprid
Mass percent be respectively 0.8%, 0.9%, 2.0%, 6.0%, 7.0%.
For the frequency domain spectra of 0.5-2.0THz characteristic wave bands, training set sample frequency domain is divided into point-score using Latin
Spectrum and checking collect sample frequency domain spectra, and N values take 4 here；Then 15 samples are randomly divided into 4 groups, each partition ensures each
Sample only collects as 1 checking, and each sample repeats partition 10 times as training set 3 times；
Wherein, the Latin matches somebody with somebody point-score, and specific calculating process is described as follows：
(a) assume that data set has n sample, sample is randomly ordered；
(b) data set is divided into the m groups (m=n/N) of equal sizes；
(c) using first group as checking collection, remaining m-1 groups are training set, i.e., with n-N training set Sample Establishing analysis mould
Type, and predict the N number of sample for being extracted and；
(d) using second group as checking collection, remaining m-1 groups are training set, i.e., with n-N training set Sample Establishing analysis mould
Type, and predict the N number of sample for being extracted and；
(e) by that analogy, carry out m times altogether, until all groups are extracted once and carry out being predicted as stopping；
(5) fixed is set up using partial least-square regression method to the training set sample and the checking collection sample frequency domain spectra
Analysis model is measured, its concrete principle is described as follows：
Deflected secondary air is built upon the model on the basis of independent variable X and dependent variable Y matrixes, by setting up independent variable
Linear regression model (LRM) of the latent variable on the latent variable of dependent variable, and then react the relation between independent variable and dependent variable；
X=TPT+ E=∑s tapa t；
Y=UQT+ F=∑s uaqa t；
In formula：T is X score matrix, and P is X loading matrix, and E is X residual matrix, taFor score vector, paFor phase
The load vectors answered, U is Y score matrix, and Q is Y loading matrix, and F is Y residual matrix, uaFor score vector, qaFor phase
The load vectors answered；
PLS extracts respective latent variable in X and Y respectively, and they are respectively independent variable and dependent variable
Linear combination, while following condition should be met：
(a) two groups of latent variables farthest carry the variation information of independent variable and dependent variable respectively；
(b) covariance therebetween is maximized；
PLS is a method with iterative method progressively extract component.The utilization pair mutually in iterative calculation
The information of side, iteration is constantly according to X, Y remaining information adjustment t each timea、uaThe constituents extraction of the second wheel is carried out, until remnants
Element absolute value in matrix is approximately zero, and algorithm stops.Thus obtained coefficient can preferably reflect X and Y relation.
The prediction mean square error of the analysis model and the graph of a relation of number of principal components are illustrated in figure 7, is selected to lead according to Fig. 7
Component number is 4.Institute's established model is verified with point-score with Latin, correction root-mean-square error (RMSEC), prediction root mean square are missed
The foundation that poor (RMSEP), prediction related coefficient (Rv) are judged as model performance, as a result as shown in table 2.
The graph of a relation between the predicted value of the analysis model and experiment value is illustrated in figure 8, illustrates the predicted value of model
With experiment value very close to prediction related coefficient can reach 0.959.
Table 2- PLSs set up the results of property of analysis model
The data from table 2 are again it can be seen that RMSEC, RMSEP of above-mentioned analysis model are smaller, RvIt is larger, so as to say
The model that bright the inventive method is set up is reliably feasible, can be for quantitatively being detected to sevin in small rice sample.
Embodiment 3
The present embodiment provides one kind and quantitatively detects imidacloprid content in rice sample using Terahertz frequency domain spectra, wherein, institute
Stating " rice sample " to be measured, (Zhangjiakou City inspection and quarantine bureau of Hebei province carries using the rice for itself not containing any residues of pesticides
For) prepare after addition known quantity imidacloprid (offers of Beijing North Na Chuanlian Bioteknologisk Institut), and will it is described it is to be measured " greatly
Rice sample " is detected as blind sample using the inventive method, is comprised the following steps that：
(1) rice is put into pulverizer and crushed, sieved (100 mesh, particle diameter≤150 μm) after grinding, be put into baking oven
Middle drying, obtains rice powder；Rice powder and imidacloprid powder are mixed according to different quality ratio, agate mortar is transferred to
In further grinding obtain " rice sample " fine powder to be measured, wherein, imidacloprid quality hundred in " rice sample " to be measured
Point content is followed successively by 0%, 0.99%, 1.5%, 1.97%, 2.50%, 3.00%, 3.50%, 4.00%, 4.50%,
5.01%, 5.50%, 5.91%, 6.51%, 7.00%, 7.50%, 8.00%, 8.51%, 9.00%, 9.50%,
10.00%, 11.99%, 14.02%, 15.01%；
Weigh in above-mentioned every kind of " rice sample " fine powder about 170mg, the mould for being placed in Specac companies, use tablet press machine
3-4min is kept under 5t pressure, so that " rice sample " fine powder by described in is pressed into diameter 13mm, thickness about 1.0mm circle
Thin slice, produces the grain press sheet compression to be measured containing different quality percentage imidacloprid, the table of grain press sheet compression two to be measured
Face is parallel, surface is smooth and no crack.
(2) by imidacloprid powder and polyethylene powders according to mass ratio 1:4 mixing, are transferred in agate mortar and further grind
Mill, the mixed pressuring plate sample is obtained using with the compacting of step (1) identical method；
(3) the transmission-type terahertz time-domain spectroscopy system based on photoconductive antenna is applied, using U.S. Spectra-
The Mai Tai types femtosecond laser oscillators of Physics (spectrum-physics) brand are as external lasing light emitter, cardiac wave in laser pulse
Long 800nm, power output ＞ 500mW, under the conditions of 20 DEG C, Terahertz frequency range is 0.3-3.0THz, with nitrogen as ginseng
Examine, using transmission measurement pattern, the terahertz time-domain spectroscopy signal for gathering the grain press sheet compression to be measured is used as sample signal
Esam(t), and before the time-domain spectroscopy data in Terahertz region of each press sheet compression are gathered, first gather without placement tabletting
The time-domain spectroscopy signal in the Terahertz region under the nitrogen atmosphere of sample is used as reference signal Eref(t)；By reference signal Eref(t)
With sample signal Esam(t) respectively through the frequency-region signal E referred to after Fast Fourier Transform (FFT)ref(ω) and sample frequency domain
Signal Esam(ω)。
When carrying out above-mentioned data acquisition, the sample spectra of each grain press sheet compression to be measured is heavy under identical environment
Second mining collection 3 times, the average value for finally taking 3 times finally gives 23 groups of frequency domain spectra numbers as the frequency domain modal data used in subsequent treatment
According to.
(4) it is illustrated in figure 9 the Terahertz frequency domain spectra that imidacloprid-polyethylene is determined using the inventive method, it can be seen that
It is the frequency domain spectra characteristic peak that there is imidacloprid at 0.89THz, 1.26THz, 1.50THz in frequency.
The grain samples to be measured for the different imidacloprid weight/mass percentage compositions described in this patent are having as shown in Figure 10
The interval frequency domain spectra of frequency range is imitated, in 0.1-2.0THz frequency range, most frequency domain spectra letters of pesticide imidacloprid are contained
Breath, therefore effective frequency range using Terahertz frequency domain spectra technical Analysis rice Pesticides imidacloprid can be used as.In addition, from figure
10 can be seen that in the range of imidacloprid mass fraction is 0.00%-15.01%, the frequency domain spectra characteristic peak and pyrrole of imidacloprid
The mass fraction of worm quinoline change in direct ratio, so the frequency domain spectra in the characteristic spectra of imidacloprid can be used in grain Pesticides pyrrole
The quantitative analysis of worm quinoline.
In the validity feature wave band of the selected grain press sheet compression to be measured, by the grain press sheet compression to be measured
Frequency domain spectra, above-mentioned 23 groups of frequency domain modal datas are divided into training set sample frequency domain spectra and checking collection sample using self-service Latin with point-score
This frequency domain spectra, selection takes 4 with fraction N, takes wherein 3/4 as training set sample, 1/4 as checking collection sample, is specially：First
Sample to be tested is divided into 4 parts, selects wherein 1 part as checking collection sample, remaining 3 parts as training set sample, it is necessary to explanation
It is that in each calculate, each sample is only used for once predicting checking.
(5) training set data collection is utilized, quantitative calibration models is set up with reference to PLS, is intersected using leaving-one method
Checking, determines the main cause subnumber of PLS, as shown in Figure 11, and when main gene is 6, the root mean square of cross validation is missed
It is poor minimum, RMSECV=0.00418 (MSECV=1.747 × 10-5), therefore use 6 main causes when setting up quantitative calibration models
Son.The concrete principle of the PLS is as follows：
PLS is decomposed to Terahertz the frequency domain spectrum matrix X and concentration matrix Y of sample first, its model
It is expressed as follows：
In above-mentioned expression formula, tk(n × 1) is the score of matrix X i-th of main gene；pk(1 × m) is i-th of matrix X
The load of main gene；uk(n × 1) is the score of concentration matrix Y i-th of main gene, qk(1 × p) is i-th of concentration matrix Y
The load of main gene；K is main gene number.T and U are the score matrix of X and Y matrixes respectively, and P and Q are the load of X and Y matrixes respectively
Lotus matrix, Ex and EYIt is then the PLS regression criterion matrixes of X and Y matrixes respectively.
T and U two score matrixes T and U are done into linear regression afterwards：
U=TB
B=(TTT)-1TTY
Finally when being predicted, obtaining for unknown sample frequency domain spectrum matrix X ＇ is obtained according to matrix X loading matrix P first
Sub-matrix T ＇, then can be obtained the concentration prediction matrix Y ＇ of unknown sample by following formula：
Y ＇=T ＇ BQ
It is hereby achieved that the concentration prediction value y ＇ of the unknown sample in concentration prediction matrix Y ＇；
(5) using checking collection data set, under optimal main cause subnumber, the partially minimum second metering straightening die set up is verified
The estimated performance of type, the actual concentrations of different samples and the comparing result of prediction concentrations as shown in table 3, as a result show：In 0-
In the range of 15% imidacloprid mass fraction, the prediction mean square error of partial least square model is MSEV=2.0561 × 10-5, put down
Square coefficient R2=0.9870.It is as shown in figure 11 graph of a relation between the model prediction concentration and actual concentrations, can be with
The correlation being apparent between the model prediction concentration and actual concentrations is satisfactory.
Table 3- difference sample actual concentrations values and prediction concentrations value
As can be seen here, the MSEV of above-mentioned analysis model prediction is smaller, R2Up to 0.9870, so as to illustrate that the inventive method is built
Vertical model is reliably feasible, can be for quantitatively being detected to rice sample Pesticides imidacloprid.
Using the method for foregoing description, the sevin in the Acetamiprid and rice in wheat, imidacloprid can not only be entered
The quantitative detection of row, can also be generalized to the grain samples of other species and the detection of other agricultural chemicals.In the sample of other species grains
In the detection of product, also need sample preparation into pressed powder.The agricultural chemicals detected should have feature to inhale in Terahertz frequency range
Receive.
Detection method of the present invention can accurately, it is easy farm chemical ingredients in grain are detected, the studies above item
Mesh obtains the subsidy of state natural sciences fund (21275101) and national great scientific instrument special (2012YQ140005).
Obviously, above-described embodiment is only intended to clearly illustrate example, and the not restriction to embodiment.It is right
For those of ordinary skill in the art, can also make on the basis of the above description it is other it is various forms of change or
Change.There is no necessity and possibility to exhaust all the enbodiments.And the obvious change thus extended out or
Among changing still in the protection domain of the invention.
Claims (3)
1. a kind of method that utilization THz-TDS frequency domain spectras quantitatively detect Pesticide Residues In Grain, it is characterised in that including following step
Suddenly：
(1) tabletting after grain samples grinding to be measured is taken, grain press sheet compression to be measured is produced；
(2) the target pesticide sample to be detected is taken, according to mass ratio 1:4 ratio is mixed with polyethylene, and is pressed after grinding
Piece, produces the mixed pressuring plate sample containing target agricultural chemicals and polyethylene；
(3) the grain press sheet compression to be measured and the mixed pressuring plate sample are surveyed one by one using THz-TDS spectroscopic systems
Examination, obtains the terahertz time-domain spectroscopy signal under the terahertz time-domain spectroscopy of each sample, collection nitrogen atmosphere as reference signal,
And the terahertz time-domain spectroscopy signal conduct of the grain press sheet compression to be measured and mixed pressuring plate sample is gathered under the same conditions
Sample signal, by reference signal and sample signal respectively through the frequency domain spectrum signal Es and sample referred to after Fourier transformation
Frequency domain spectrum signal Er：
In formula, ω represents frequency,The phase difference of representative sample signal and reference signal, ρ is shaking for sample signal and reference signal
The ratio of width mould；
(4) according to the frequency domain spectrum signal of the mixed pressuring plate sample, the preferable wave band interval of reappearance is selected as characteristic wave bands,
And by frequency domain spectra random division of the grain press sheet compression to be measured under the characteristic wave bands be training set sample frequency domain spectra and
Checking collection sample frequency domain spectra；
(5) the training set sample frequency domain spectra and the checking collection sample frequency domain spectra are set up using partial least-square regression method
Quantitative Analysis Model, obtains the quantitative detected value of each grain samples to be measured；
The agricultural chemicals is Acetamiprid, and the Acetamiprid frequency domain spectra characteristic wave bands are 0.1-2.0THz；
In the step (5), the training set sample frequency domain spectra and the checking collection are set up using partial least-square regression method
The Quantitative Analysis Model of sample frequency domain spectra, the number of principal components of the Acetamiprid frequency domain spectra Quantitative Analysis Model is 2；
In the step (3), the test condition of the THz-TDS is：Detection temperature is 19 DEG C, with nitrogen as reference, spectrum
The scanning step motor interval of instrument scanning system is -1mm-2mm, and step-length is 0.01mm；
In the step (4), the division of the training set sample and the checking collection sample is carried out using self-service Latin with point-score；
Specific calculating process of the self-service Latin with point-score is as follows：
(a) assume that data set has n sample, sample is randomly ordered, extract N number of sample out from data set every time afterwards；
(b) data set is divided into the m groups of equal sizes, m=n/N；
(c) using first group as checking collection, remaining m-1 groups are training set, i.e., with n-N training set Sample Establishing analysis model, and
Prediction is extracted the N number of sample come；
(d) using second group as checking collection, remaining m-1 groups are training set, i.e., with n-N training set Sample Establishing analysis model, and
Prediction is extracted the N number of sample come；
(e) by that analogy, carry out m times altogether, until all groups are extracted once and carry out being predicted as stopping；
When being divided using the self-service Latin with point-score to the frequency domain spectra, repeat partition and calculate 10 times；
In the step (3), each sample duplicate measurements 3 times, each sample is diameter 13mm, thickness 1.2mm circle sheet；
In the step (5), the specific Computing Principle of the partial least-square regression method is as follows：
Deflected secondary air is built upon the model on the basis of independent variable X and dependent variable Y matrixes, by setting up the latent of independent variable
Linear regression model (LRM) of the variable on the latent variable of dependent variable, and then react the relation between independent variable and dependent variable；
X=TPT+ E=∑s tapa t；
Y=UQT+ F=∑s uaqa t；
In formula：T is X score matrix, and P is X loading matrix, and E is X residual matrix, taFor score vector, paTo be corresponding
Load vectors, U is Y score matrix, and Q is Y loading matrix, and F is Y residual matrix, uaFor score vector, qaTo be corresponding
Load vectors；
PLS extracts respective latent variable in X and Y respectively, and they are respectively the linear of independent variable and dependent variable
Combination, while following condition should be met：
(a) two groups of latent variables farthest carry the variation information of independent variable and dependent variable respectively；
(b) covariance therebetween is maximized.
2. a kind of method that utilization THz-TDS frequency domain spectras quantitatively detect Pesticide Residues In Grain, it is characterised in that including following step
Suddenly：
(1) tabletting after grain samples grinding to be measured is taken, grain press sheet compression to be measured is produced；
(2) the target pesticide sample to be detected is taken, according to mass ratio 1:4 ratio is mixed with polyethylene, and is pressed after grinding
Piece, produces the mixed pressuring plate sample containing target agricultural chemicals and polyethylene；
(3) the grain press sheet compression to be measured and the mixed pressuring plate sample are surveyed one by one using THz-TDS spectroscopic systems
Examination, obtains the terahertz time-domain spectroscopy signal under the terahertz time-domain spectroscopy of each sample, collection nitrogen atmosphere as reference signal,
And the terahertz time-domain spectroscopy signal conduct of the grain press sheet compression to be measured and mixed pressuring plate sample is gathered under the same conditions
Sample signal, by reference signal and sample signal respectively through the frequency domain spectrum signal Es and sample referred to after Fourier transformation
Frequency domain spectrum signal Er：
In formula, ω represents frequency,The phase difference of representative sample signal and reference signal, ρ is shaking for sample signal and reference signal
The ratio of width mould；
(4) according to the frequency domain spectrum signal of the mixed pressuring plate sample, the preferable wave band interval of reappearance is selected as characteristic wave bands,
And by frequency domain spectra random division of the grain press sheet compression to be measured under the characteristic wave bands be training set sample frequency domain spectra and
Checking collection sample frequency domain spectra；
(5) the training set sample frequency domain spectra and the checking collection sample frequency domain spectra are set up using partial least-square regression method
Quantitative Analysis Model, obtains the quantitative detected value of each grain samples to be measured；
The agricultural chemicals is sevin, and the frequency domain spectra characteristic wave bands of the sevin are 0.5-2.0THz；
In the step (5), the training set sample frequency domain spectra and the checking collection are set up using partial least-square regression method
The Quantitative Analysis Model of sample frequency domain spectra, the number of principal components of the sevin frequency domain spectra Quantitative Analysis Model is 4；
In the step (3), the test condition of the THz-TDS is：Detection temperature is 19 DEG C, with nitrogen as reference, spectrum
The scanning step motor interval of instrument scanning system is -1mm-2mm, and step-length is 0.01mm；
In the step (4), the division of the training set sample and the checking collection sample is carried out using self-service Latin with point-score；
Specific calculating process of the self-service Latin with point-score is as follows：
(a) assume that data set has n sample, sample is randomly ordered, extract N number of sample out from data set every time afterwards；
(b) data set is divided into the m groups of equal sizes, m=n/N；
(c) using first group as checking collection, remaining m-1 groups are training set, i.e., with n-N training set Sample Establishing analysis model, and
Prediction is extracted the N number of sample come；
(d) using second group as checking collection, remaining m-1 groups are training set, i.e., with n-N training set Sample Establishing analysis model, and
Prediction is extracted the N number of sample come；
(e) by that analogy, carry out m times altogether, until all groups are extracted once and carry out being predicted as stopping；
When being divided using the self-service Latin with point-score to the frequency domain spectra, repeat partition and calculate 10 times；
In the step (3), each sample duplicate measurements 3 times, each sample is diameter 13mm, thickness 1.2mm circle sheet；
In the step (5), the specific Computing Principle of the partial least-square regression method is as follows：
Deflected secondary air is built upon the model on the basis of independent variable X and dependent variable Y matrixes, by setting up the latent of independent variable
Linear regression model (LRM) of the variable on the latent variable of dependent variable, and then react the relation between independent variable and dependent variable；
X=TPT+ E=∑s tapa t；
Y=UQT+ F=∑s uaqa t；
In formula：T is X score matrix, and P is X loading matrix, and E is X residual matrix, taFor score vector, paTo be corresponding
Load vectors, U is Y score matrix, and Q is Y loading matrix, and F is Y residual matrix, uaFor score vector, qaTo be corresponding
Load vectors；
PLS extracts respective latent variable in X and Y respectively, and they are respectively the linear of independent variable and dependent variable
Combination, while following condition should be met：
(a) two groups of latent variables farthest carry the variation information of independent variable and dependent variable respectively；
(b) covariance therebetween is maximized.
3. a kind of method that utilization THz-TDS frequency domain spectras quantitatively detect Pesticide Residues In Grain, it is characterised in that including following step
Suddenly：
(1) tabletting after grain samples grinding to be measured is taken, grain press sheet compression to be measured is produced；
(2) the target pesticide sample to be detected is taken, according to mass ratio 1:4 ratio is mixed with polyethylene, and is pressed after grinding
Piece, produces the mixed pressuring plate sample containing target agricultural chemicals and polyethylene；
(3) the grain press sheet compression to be measured and the mixed pressuring plate sample are surveyed one by one using THz-TDS spectroscopic systems
Examination, obtains the terahertz time-domain spectroscopy signal under the terahertz time-domain spectroscopy of each sample, collection nitrogen atmosphere as reference signal,
And the terahertz time-domain spectroscopy signal conduct of the grain press sheet compression to be measured and mixed pressuring plate sample is gathered under the same conditions
Sample signal, by reference signal and sample signal respectively through the frequency domain spectrum signal Es and sample referred to after Fourier transformation
Frequency domain spectrum signal Er：
In formula, ω represents frequency,The phase difference of representative sample signal and reference signal, ρ is shaking for sample signal and reference signal
The ratio of width mould；
(4) according to the frequency domain spectrum signal of the mixed pressuring plate sample, the preferable wave band interval of reappearance is selected as characteristic wave bands,
And by frequency domain spectra random division of the grain press sheet compression to be measured under the characteristic wave bands be training set sample frequency domain spectra and
Checking collection sample frequency domain spectra；
(5) the training set sample frequency domain spectra and the checking collection sample frequency domain spectra are set up using partial least-square regression method
Quantitative Analysis Model, obtains the quantitative detected value of each grain samples to be measured；
The agricultural chemicals is imidacloprid, and the frequency domain spectra characteristic wave bands of the imidacloprid are in 0.1-2.0THz；
In the step (5), the training set sample frequency domain spectra and the checking collection are set up using partial least-square regression method
The Quantitative Analysis Model of sample frequency domain spectra, the main cause subnumber of the imidacloprid frequency domain spectra Quantitative Analysis Model is 6；
In the step (3), the test condition of the THz-TDS is：Detection temperature is 19 DEG C, with nitrogen as reference, spectrum
The scanning step motor interval of instrument scanning system is -1mm-2mm, and step-length is 0.01mm；
In the step (4), the division of the training set sample and the checking collection sample is carried out using self-service Latin with point-score；
Specific calculating process of the self-service Latin with point-score is as follows：
(a) assume that data set has n sample, sample is randomly ordered, extract N number of sample out from data set every time afterwards；
(b) data set is divided into the m groups of equal sizes, m=n/N；
(c) using first group as checking collection, remaining m-1 groups are training set, i.e., with n-N training set Sample Establishing analysis model, and
Prediction is extracted the N number of sample come；
(d) using second group as checking collection, remaining m-1 groups are training set, i.e., with n-N training set Sample Establishing analysis model, and
Prediction is extracted the N number of sample come；
(e) by that analogy, carry out m times altogether, until all groups are extracted once and carry out being predicted as stopping；
When being divided using the self-service Latin with point-score to the frequency domain spectra, repeat partition and calculate 10 times；
In the step (3), each sample duplicate measurements 3 times, each sample is diameter 13mm, thickness 1.2mm circle sheet；
In the step (5), the specific Computing Principle of the partial least-square regression method is as follows：
Deflected secondary air is built upon the model on the basis of independent variable X and dependent variable Y matrixes, by setting up the latent of independent variable
Linear regression model (LRM) of the variable on the latent variable of dependent variable, and then react the relation between independent variable and dependent variable；
X=TPT+ E=∑s tapa t；
Y=UQT+ F=∑s uaqa t；
In formula：T is X score matrix, and P is X loading matrix, and E is X residual matrix, taFor score vector, paTo be corresponding
Load vectors, U is Y score matrix, and Q is Y loading matrix, and F is Y residual matrix, uaFor score vector, qaTo be corresponding
Load vectors；
PLS extracts respective latent variable in X and Y respectively, and they are respectively the linear of independent variable and dependent variable
Combination, while following condition should be met：
(a) two groups of latent variables farthest carry the variation information of independent variable and dependent variable respectively；
(b) covariance therebetween is maximized.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201410510382.2A CN104297202B (en) | 2014-09-28 | 2014-09-28 | The method for quantitatively detecting Pesticide Residues In Grain using THz TDS frequency domain spectras |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201410510382.2A CN104297202B (en) | 2014-09-28 | 2014-09-28 | The method for quantitatively detecting Pesticide Residues In Grain using THz TDS frequency domain spectras |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN104297202A CN104297202A (en) | 2015-01-21 |
| CN104297202B true CN104297202B (en) | 2017-10-20 |
Family
ID=52317026
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201410510382.2A Active CN104297202B (en) | 2014-09-28 | 2014-09-28 | The method for quantitatively detecting Pesticide Residues In Grain using THz TDS frequency domain spectras |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN104297202B (en) |
Families Citing this family (4)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN107290302A (en) * | 2017-07-12 | 2017-10-24 | 中央民族大学 | Chromium quantitative detecting method in edible gelatin based on terahertz time-domain spectroscopy |
| CN107727608A (en) * | 2017-11-17 | 2018-02-23 | 深圳市太赫兹系统设备有限公司 | The detection method of Indomethacin |
| CN109470647B (en) * | 2019-01-20 | 2020-11-24 | 南京林业大学 | Method for measuring terahertz absorption spectrum of water vapor |
| CN114264627A (en) * | 2021-12-20 | 2022-04-01 | 河南工业大学 | Terahertz sensor and using method thereof |
Family Cites Families (5)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| US7605371B2 (en) * | 2005-03-01 | 2009-10-20 | Osaka University | High-resolution high-speed terahertz spectrometer |
| JP2010038809A (en) * | 2008-08-07 | 2010-02-18 | Murata Mfg Co Ltd | Terahertz spectroscopic device |
| CN103411912A (en) * | 2013-05-22 | 2013-11-27 | 首都师范大学 | Method for identifying Chinese herbal medicine by using THz-TDS (terahertz-total dissolved solids) in combination with fuzzy rule expert system |
| CN103424372A (en) * | 2013-08-26 | 2013-12-04 | 青岛科技大学 | Solid pesticide classification method based on decision tree algorithm |
| CN103969214B (en) * | 2014-04-17 | 2016-08-31 | 首都师范大学 | The method utilizing Terahertz frequency range infrared spectrum technology detection grain Pesticides content |
-
2014
- 2014-09-28 CN CN201410510382.2A patent/CN104297202B/en active Active
Also Published As
| Publication number | Publication date |
|---|---|
| CN104297202A (en) | 2015-01-21 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN103969214B (en) | The method utilizing Terahertz frequency range infrared spectrum technology detection grain Pesticides content | |
| CN104181122A (en) | Method for detecting content of pesticide in cereal by utilizing terahertz time-domain spectroscopy | |
| Hua et al. | Qualitative and quantitative detection of pesticides with terahertz time-domain spectroscopy | |
| CN103969212B (en) | The method utilizing Terahertz frequency range FTIR technology detection by quantitative Pesticide Residues In Grain | |
| CN104215599A (en) | Method for quantitatively detecting pesticide residue in grain through THz-TDS technology | |
| CN104237157A (en) | Method for utilizing terahertz time-domain spectroscopy technology to detect amino acid content in grain | |
| CN104297202B (en) | The method for quantitatively detecting Pesticide Residues In Grain using THz TDS frequency domain spectras | |
| CN104330384B (en) | Using the method for Terahertz frequency domain spectra technology for detection Amino-Acid in Grain content | |
| CN104406923A (en) | Method for quantitative detection of amino acid content in grain by THz-TDS technology | |
| Chia et al. | Neural network and principal component regression in non-destructive soluble solids content assessment: a comparison | |
| Yin et al. | Identification of edible oils using terahertz spectroscopy combined with genetic algorithm and partial least squares discriminant analysis | |
| CN104034689A (en) | Method and device for detecting heavy metal content of soil on basis of tabletting method | |
| CN104849233A (en) | Cereal freshness detecting method and device | |
| CN106525761A (en) | Nitrite detection method based on terahertz spectroscopy scanning | |
| Lu et al. | Detection of heavy metals in vegetable soil based on THz spectroscopy | |
| CN108956584B (en) | The quick and precisely detection method of heavy metal element chromium in a kind of mulberry fruit | |
| CN104596979A (en) | Method for measuring cellulose of reconstituted tobacco by virtue of near infrared reflectance spectroscopy technique | |
| CN111965134A (en) | Terahertz spectrum quantitative analysis method for rubber vulcanization accelerator mixture | |
| Lian et al. | Qualitative and quantitative analyses of potassium sorbate in milk powder using terahertz spectra | |
| Sun et al. | Nondestructive detection of melamine in milk powder by terahertz spectroscopy and correlation analysis algorithm | |
| CN105866043A (en) | Method for detecting apple sour through hyperspectral technology | |
| CN108982407A (en) | A method of probing into the soil optimum moisture content of detection soil nitrogen using near infrared spectrum | |
| Jiang et al. | Determination of potassium sorbate and sorbic acid in agricultural products using THz time-domain spectroscopy | |
| Wentao et al. | Study on identification methods in the detection of transgenic material based on terahertz time domain spectroscopy | |
| Zhong et al. | THz wave standoff detection of explosive materials |
Legal Events
| Date | Code | Title | Description |
|---|---|---|---|
| C06 | Publication | ||
| PB01 | Publication | ||
| C10 | Entry into substantive examination | ||
| SE01 | Entry into force of request for substantive examination | ||
| GR01 | Patent grant | ||
| GR01 | Patent grant |