CN109444066B - Model transfer method based on spectral data - Google Patents
Model transfer method based on spectral data Download PDFInfo
- Publication number
- CN109444066B CN109444066B CN201811271230.6A CN201811271230A CN109444066B CN 109444066 B CN109444066 B CN 109444066B CN 201811271230 A CN201811271230 A CN 201811271230A CN 109444066 B CN109444066 B CN 109444066B
- Authority
- CN
- China
- Prior art keywords
- instrument
- spectrum
- sample
- samples
- transfer
- 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
- 238000012546 transfer Methods 0.000 title claims abstract description 162
- 238000000034 method Methods 0.000 title claims abstract description 105
- 230000003595 spectral effect Effects 0.000 title claims abstract description 90
- 238000001228 spectrum Methods 0.000 claims abstract description 152
- 238000012937 correction Methods 0.000 claims abstract description 87
- 230000002441 reversible effect Effects 0.000 claims abstract description 8
- 238000005259 measurement Methods 0.000 claims abstract description 5
- 238000002329 infrared spectrum Methods 0.000 claims description 37
- 239000011159 matrix material Substances 0.000 claims description 35
- 238000006243 chemical reaction Methods 0.000 claims description 15
- 230000009466 transformation Effects 0.000 claims description 10
- 238000004164 analytical calibration Methods 0.000 claims description 8
- 238000004364 calculation method Methods 0.000 claims description 8
- 238000000354 decomposition reaction Methods 0.000 claims description 7
- 238000011156 evaluation Methods 0.000 claims description 6
- 230000008569 process Effects 0.000 claims description 5
- 230000001419 dependent effect Effects 0.000 claims description 4
- 238000001514 detection method Methods 0.000 claims description 4
- 230000007704 transition Effects 0.000 claims description 3
- 238000004566 IR spectroscopy Methods 0.000 claims description 2
- 238000005481 NMR spectroscopy Methods 0.000 claims description 2
- 238000001069 Raman spectroscopy Methods 0.000 claims description 2
- 238000009499 grossing Methods 0.000 claims description 2
- 238000010606 normalization Methods 0.000 claims description 2
- 238000000870 ultraviolet spectroscopy Methods 0.000 claims description 2
- 238000007781 pre-processing Methods 0.000 abstract description 5
- 230000002457 bidirectional effect Effects 0.000 abstract description 3
- 230000000694 effects Effects 0.000 description 15
- 239000008186 active pharmaceutical agent Substances 0.000 description 5
- 238000002835 absorbance Methods 0.000 description 4
- 230000008859 change Effects 0.000 description 4
- 239000000284 extract Substances 0.000 description 4
- SMDOOINVMJSDPS-UHFFFAOYSA-N Astragaloside Natural products C1=C(O)C(OC)=CC(C2=C(C(=O)C3=C(O)C=C(O)C=C3O2)OC2C(C(OC3C(C(O)C(O)C(CO)O3)O)C(O)C(CO)O2)O)=C1 SMDOOINVMJSDPS-UHFFFAOYSA-N 0.000 description 3
- 230000002159 abnormal effect Effects 0.000 description 3
- QMNWISYXSJWHRY-XWJCTJPOSA-N astragaloside Chemical compound O1[C@H](C(C)(O)C)CC[C@]1(C)[C@@H]1[C@@]2(C)CC[C@]34C[C@]4(CC[C@H](O[C@H]4[C@@H]([C@@H](O)[C@H](O)CO4)O)C4(C)C)C4[C@@H](O[C@H]4[C@@H]([C@@H](O)[C@H](O)[C@@H](CO)O4)O)CC3[C@]2(C)C[C@@H]1O QMNWISYXSJWHRY-XWJCTJPOSA-N 0.000 description 3
- 238000010183 spectrum analysis Methods 0.000 description 3
- 238000010561 standard procedure Methods 0.000 description 3
- 239000013598 vector Substances 0.000 description 3
- 241000045403 Astragalus propinquus Species 0.000 description 2
- 239000009636 Huang Qi Substances 0.000 description 2
- 238000004458 analytical method Methods 0.000 description 2
- QMNWISYXSJWHRY-YLNUDOOFSA-N astragaloside IV Chemical compound O1[C@H](C(C)(O)C)CC[C@]1(C)[C@@H]1[C@@]2(C)CC[C@]34C[C@]4(CC[C@H](O[C@H]4[C@@H]([C@@H](O)[C@H](O)CO4)O)C4(C)C)[C@H]4[C@@H](O[C@H]4[C@@H]([C@@H](O)[C@H](O)[C@@H](CO)O4)O)C[C@H]3[C@]2(C)C[C@@H]1O QMNWISYXSJWHRY-YLNUDOOFSA-N 0.000 description 2
- QMNWISYXSJWHRY-BCBPIKMJSA-N astragaloside IV Natural products CC(C)(O)[C@@H]1CC[C@@](C)(O1)[C@H]2[C@@H](O)C[C@@]3(C)[C@@H]4C[C@H](O[C@@H]5O[C@H](CO)[C@H](O)[C@@H](O)[C@H]5O)[C@H]6C(C)(C)[C@H](CC[C@@]67C[C@@]47CC[C@]23C)O[C@@H]8OC[C@@H](O)[C@H](O)[C@H]8O QMNWISYXSJWHRY-BCBPIKMJSA-N 0.000 description 2
- 235000006533 astragalus Nutrition 0.000 description 2
- 230000005540 biological transmission Effects 0.000 description 2
- PFKIBRPYVNVMRU-UHFFFAOYSA-N cyclosieversioside F Natural products CC(C)(O)C1COC(C)(C1)C2C(O)CC3(C)C4CC(OC5OC(CO)C(O)C(O)C5O)C6C(C)(C)C(CCC67CC47CCC23C)OC8OCC(O)C(O)C8O PFKIBRPYVNVMRU-UHFFFAOYSA-N 0.000 description 2
- 238000001914 filtration Methods 0.000 description 2
- 230000002401 inhibitory effect Effects 0.000 description 2
- 239000000463 material Substances 0.000 description 2
- 230000004048 modification Effects 0.000 description 2
- 238000012986 modification Methods 0.000 description 2
- 230000036961 partial effect Effects 0.000 description 2
- 239000000126 substance Substances 0.000 description 2
- 238000004497 NIR spectroscopy Methods 0.000 description 1
- 230000032683 aging Effects 0.000 description 1
- 235000019206 astragalus extract Nutrition 0.000 description 1
- 230000009286 beneficial effect Effects 0.000 description 1
- 238000006555 catalytic reaction Methods 0.000 description 1
- 230000000052 comparative effect Effects 0.000 description 1
- 230000007423 decrease Effects 0.000 description 1
- 230000007547 defect Effects 0.000 description 1
- 238000005516 engineering process Methods 0.000 description 1
- 238000004128 high performance liquid chromatography Methods 0.000 description 1
- 230000006872 improvement Effects 0.000 description 1
- 238000012844 infrared spectroscopy analysis Methods 0.000 description 1
- 230000000670 limiting effect Effects 0.000 description 1
- 238000004811 liquid chromatography Methods 0.000 description 1
- 238000010238 partial least squares regression Methods 0.000 description 1
- 238000012628 principal component regression Methods 0.000 description 1
- 230000004044 response Effects 0.000 description 1
- 238000012552 review Methods 0.000 description 1
- 238000000638 solvent extraction Methods 0.000 description 1
- 238000004611 spectroscopical analysis Methods 0.000 description 1
- 239000004575 stone Substances 0.000 description 1
- 230000001131 transforming effect Effects 0.000 description 1
Images
Classifications
-
- G—PHYSICS
- G01—MEASURING; TESTING
- G01N—INVESTIGATING OR ANALYSING MATERIALS BY DETERMINING THEIR CHEMICAL OR PHYSICAL PROPERTIES
- G01N21/00—Investigating or analysing materials by the use of optical means, i.e. using sub-millimetre waves, infrared, visible or ultraviolet light
- G01N21/17—Systems in which incident light is modified in accordance with the properties of the material investigated
- G01N21/25—Colour; Spectral properties, i.e. comparison of effect of material on the light at two or more different wavelengths or wavelength bands
- G01N21/31—Investigating relative effect of material at wavelengths characteristic of specific elements or molecules, e.g. atomic absorption spectrometry
- G01N21/35—Investigating relative effect of material at wavelengths characteristic of specific elements or molecules, e.g. atomic absorption spectrometry using infrared light
-
- G—PHYSICS
- G01—MEASURING; TESTING
- G01N—INVESTIGATING OR ANALYSING MATERIALS BY DETERMINING THEIR CHEMICAL OR PHYSICAL PROPERTIES
- G01N21/00—Investigating or analysing materials by the use of optical means, i.e. using sub-millimetre waves, infrared, visible or ultraviolet light
- G01N21/17—Systems in which incident light is modified in accordance with the properties of the material investigated
- G01N21/25—Colour; Spectral properties, i.e. comparison of effect of material on the light at two or more different wavelengths or wavelength bands
- G01N21/31—Investigating relative effect of material at wavelengths characteristic of specific elements or molecules, e.g. atomic absorption spectrometry
- G01N21/35—Investigating relative effect of material at wavelengths characteristic of specific elements or molecules, e.g. atomic absorption spectrometry using infrared light
- G01N21/359—Investigating relative effect of material at wavelengths characteristic of specific elements or molecules, e.g. atomic absorption spectrometry using infrared light using near infrared light
Landscapes
- Physics & Mathematics (AREA)
- Spectroscopy & Molecular Physics (AREA)
- Health & Medical Sciences (AREA)
- Life Sciences & Earth Sciences (AREA)
- Chemical & Material Sciences (AREA)
- Analytical Chemistry (AREA)
- Biochemistry (AREA)
- General Health & Medical Sciences (AREA)
- General Physics & Mathematics (AREA)
- Immunology (AREA)
- Pathology (AREA)
- Investigating Or Analysing Materials By Optical Means (AREA)
Abstract
The embodiment of the application discloses a model transfer method based on spectral data, and comprises index measurement and spectrum acquisition of a sample; preprocessing a spectrum; dividing the calibration samples into correction samples and prediction samples; carrying out model transfer from the main instrument to the target instrument, namely forward transfer; or, the target instrument carries out model transfer to the main instrument, namely, reverse transfer; compared with the traditional method, the method has the following advantages: through the correlation of the sample light spectrum data and the spectrum data structure, the interference of noise and irrelevant information can be inhibited; the method can realize the bidirectional transfer of the models of the main instrument and the target instrument, and has important significance for sharing the models of different types of spectrometers in the same spectrum method and improving the applicability of the models, wherein the types of the instruments of the main instrument and the target instrument are consistent or not, and the number of the collected spectrum data points is consistent or not.
Description
Technical Field
The embodiment of the application relates to the technical field of infrared spectroscopy analysis, in particular to a model transfer method based on spectral data.
Background
Modern spectral analysis techniques are comprehensive analysis techniques in which multiple disciplines (spectral detection, chemometrics, computer disciplines, statistics, etc.) are cross-fused. The spectral analysis technology is widely applied to the fields of petrochemical industry, agriculture, food and pharmacy. The method comprises the steps of adopting a spectral analysis method for measuring the content of a substance component and analyzing an online process, needing to establish a multivariate calibration model, namely adopting a standard method to measure indexes (such as component content or activity value or other physical and chemical indexes) of calibration samples, then adopting a spectrometer to measure the spectrums of the samples, associating the spectrum data with the index data to establish the calibration model, substituting the spectrums of unknown samples into the calibration model, calculating the predicted value of the indexes of the unknown samples, and realizing the measurement of the indexes of the unknown samples.
In practical application, a large amount of manpower and material resources are required to be invested for establishing the spectrum model, and the cost is huge. If the established model is applied to other instruments or used continuously, a large amount of manpower and material resources are saved, and the method has important significance for realizing the sharing of the model and improving the universality and the applicability of the model. With the diversification of spectrum instruments and the complexity of application environments, such as the types and models of the instruments are the same or different, the aging of the instruments and the replacement of parts, the change of temperature and humidity of a sample measuring environment and other factors, the collected spectrum often has the problems of difference of spectral absorbance (or intensity), drift of wavelength (or wave number) and the like, and if the established model is continuously used, the deviation of a prediction result is large, and even the model cannot be used. In order to solve the technical bottleneck problem of model sharing and improving the universality of the model, model transfer or transfer is an effective method. The model transfer is to realize the consistency and the accuracy of the model prediction results of the main instrument and the target instrument by establishing the functional relationship between the detection signals of the main instrument and the target instrument (or the sub-instrument). According to the associated objects of the established functions, the model transfer method is divided into three categories: one is a calibration method based on the prediction results, such as the S/B method mentioned in Tom feed, Standard disparation and calibration Transfer for near Infrared instruments, A Review, Journal of near Infrared Spectroscopy Vol.9, Issue 4, pp.229-244 (2001); second, a calibration method based on spectrum signals, such as the methods of Shenks, DS, PDS, FIR, OSC, etc.; the third is the model parameter-based calibration method mentioned in M.Forina, G.Drava, C.Armanino, R.Boggia, S.Lanteri, R.Leardi, P.Corti, P.Conti, R.Gianggiacomo, C.Galliena, R.Bigoni, I.Quartari, C.Serra, D.Ferri, O.Leoni, L.Lazzeri, Transfer of catalysis function in near-isolated spectroscopy, Chemometrics and dInterlingt Laboratory Systems, Volume 27, Issue 2, February 1995, Pages 189-203. Among them, the DS PDS method based on spectral signal correction is widely used.
Both the DS method and the PDS method belong to standard methods. The standard sample is a standard sample required for establishing a transfer function in the model transfer process. In practical applications, if a standard sample is not available or is difficult to obtain, a non-standard method, such as a Finite Impulse Response (FIR) method, may be used. The DS method is to directly correlate a main instrument and a target instrument, and then solve a transfer matrix through the least square principle to establish a functional relation. Due to the regionality of the spectral change, the spectral change may not be consistent under different wavelengths; and the noise also has regional inequality, so that the effect of the DS method for model transfer is not ideal enough. The PDS method assumes that the spectral change is mainly limited in a certain region, so that the model transfer is performed by using the local spectrum in the window, and the model transfer effect is improved to a certain extent. However, the method is complex in adjusting the window size, cannot be self-adaptive, and has a poor transfer effect and a certain influence on local noise if the window size parameter is improperly set.
Disclosure of Invention
In order to solve the defects of the prior art, the embodiment of the application provides a model transfer method based on spectral data, wherein the spectral data of a main instrument is associated with a spectral data structure of a target instrument (or the spectral data of the target instrument is associated with the spectral data structure of the main instrument), so that the method has an inhibiting effect on the interference of noise and redundant characteristics, can realize the bidirectional transmission of models of the main instrument and the target instrument, and has no limit on whether the instrument types of the main instrument and the target instrument are consistent or not and on whether the number of collected spectral data points is consistent or not; the method can convert the spectrum with lower resolution into the spectrum with higher resolution, and expands the application range of model transfer.
The embodiment of the application provides a model transfer method based on spectral data;
a method of model transfer based on spectral data, comprising:
step (1): index determination and spectrum collection of samples: aiming at a plurality of calibration samples, n indexes are measured by adopting an industry standard detection method; respectively utilizing a main instrument and a target instrument to collect the spectrums of the calibration samples to obtain the spectrum X collected by the main instrumentMAnd the spectrum X collected by the target instrumentT；
Step (2): pretreatment of the spectrum: spectrum X collected from the host instrumentMAnd the spectrum X collected by the target instrumentTRespectively adopting the same mode to carry out pretreatment;
and (3): dividing the calibration samples into correction samples and prediction samples, and acquiring the spectrum X of the main instrument according to the correction samples and the prediction samplesMDividing the samples into correction set samples and prediction set samples; spectrum X collected by target instrument according to correction sample and prediction sampleTDividing the samples into correction set samples and prediction set samples;
and (4): model transfer is carried out from the main instrument to the target instrument, namely forward transfer:
a step (401): selecting a set number of samples from the main instrument correction set samples as transfer set samples, then selecting samples with the same serial number from the target instrument correction set as target instrument transfer set samples, wherein the transfer set samples of the main instrument correspond to the serial numbers of the target instrument transfer set samples one by one; transferring the spectrum X in the sample set of the main instrumentMThe spectral variation and wavelength range of (A) are adjusted to match the spectrum X in the target instrument transfer set sampleTThe spectral variation of (a) is consistent with the wavelength range;
step (402): correlating the spectral intensity data of the main instrument transfer set sample with the spectral intensity data of the target instrument transfer set sample to obtain a relation model, and obtaining a conversion matrix;
step (403): substituting the spectral data of the main instrument correction set sample into the conversion matrix to obtain the spectral data suitable for the target instrument correction set sample;
a step (404): establishing a target instrument correction model based on the spectral data suitable for the target instrument correction set sample: the spectral data suitable for the target instrument correction set sample are used as independent variables, n measuring indexes are used as dependent variables, a multivariate correction method is adopted to correlate the spectral data suitable for the target instrument correction set sample with the n index data, and n correction models are established;
step (405): and (3) carrying out spectrum acquisition on the sample to be detected by using a target instrument, and then predicting the index value of the sample by using a target instrument correction model.
After the step (404), before the step (405), further comprising: and substituting the spectrum of the target instrument prediction set into the correction model, calculating the predicted values of the n indexes of each prediction set sample through the correction model, comparing the predicted values with the actual measured values, and performing model evaluation.
Optionally, the step (4) is replaced by: model transfer is carried out from the target instrument to the main instrument, namely reverse transfer:
step (411): selecting a set number of samples from the main instrument correction set samples as transfer set samples, then selecting samples with the same serial number from the target instrument correction set as target instrument transfer set samples, wherein the transfer set samples of the main instrument correspond to the target instrument transfer set samples one by one; transferring the spectrum X in the sample set of the main instrumentMThe spectral variation and wavelength range of (A) are adjusted to match the spectrum X in the target instrument transfer set sampleTThe spectral variation of (a) is consistent with the wavelength range;
step (412): correlating the spectral intensity data of the target instrument transfer set sample with the spectral intensity data of the main instrument transfer set sample to obtain a relation model, and obtaining a conversion matrix;
step (413): collecting the spectrum of the sample to be detected by using the target instrument, and correcting the spectrum of the sample to be detected of the target instrument into the spectrum of the sample to be detected suitable for the main instrument based on the relation model;
a step (414): establishing a correction model by using the spectrum of the main instrument correction set sample: taking spectral data of a correction set sample of a main instrument as an independent variable, taking n measurement indexes as dependent variables, and establishing n correction models by adopting a multivariate correction method; and replacing the spectrum suitable for the sample to be predicted of the main instrument into a correction model of the main instrument to predict the index value of the sample to be predicted.
After the step (414), the method may further include: and correcting the spectrum of the prediction set of the target instrument into a prediction sample spectrum suitable for the main instrument by adopting a relational model, substituting the prediction sample spectrum into a correction model, calculating the predicted values of n indexes of the samples of each prediction set through the correction model, comparing the predicted values with the actual measured values, and performing model evaluation.
Optionally, during the forward transfer, the target instrument is transferred to set spectrum XTtCarrying out main component decomposition to obtain:
XTt＝STtVTt’； (1)
wherein S isTtRepresenting a principal component score matrix, VTtRepresenting a principal component load matrix, the symbol "'" representing the transpose of the principal component load matrix, the subscript "T" representing the target instrument, and the subscript "T" representing the transfer set;
setting the number of principal components lT(ii) a Transfer set spectra from master instrument XMtAnd STtAnd (3) association:
STt＝XMtFMt； (2)
wherein, FMtRepresenting the transformation matrix, the symbol "M" representing the master, and "t" representing the transition set;
according to FMt＝X+ MtSTtCalculating and solving, wherein the symbol "+" represents generalized inverse operation; x+ MtRepresents XMtThe generalized inverse matrix of (2);
substituting the formula (2) into the formula (1) to obtain the spectrum after model conversion:
Xt T＝XMFMtVTt’； (3)
wherein, Xt TRepresenting the spectral matrix adapted to the target instrument sample after transformation, the symbol "T" representing the transformation, the symbol "T" representing the target instrument, XMRepresenting spectral data measured based on the master instrument;
the spectra of the master instrument calibration set are calibrated as follows equation (3):
Xt Tc＝XMcFMtVTt’； (4)
correcting set spectrum X of main instrument by formula (4)McCorrected to fit the target instrument spectrum Xt TcWhere the symbol "c" represents the correction set.
Optionally, the master instrument is transferred to the collection spectrum X during the reverse transfer processMtCarrying out main component decomposition to obtain:
XMt＝SMtVMt’； (5)
wherein S isMtA score matrix representing principal components, "M" representing a master instrument, "t" representing a transfer set; vMtRepresenting a principal component loading matrix, the symbol "'" representing the transpose of the matrix;
setting the number of principal components lMTransferring the target instrument to the set of spectra XTtAnd SMtAnd (3) association:
SMt＝XTtFTt； (6)
wherein, FTtRepresenting the transformation matrix, the symbol "T" representing the target instrument, "T" representing the transfer set;
wherein according to FTt＝X+ TtSMtCalculating and solving; x+ TtRepresents XTtThe generalized inverse matrix of (2);
substituting equation (6) into equation (5), model-converted spectra:
Xt M＝XTFTtVMt’； (7)
wherein, Xt MRepresenting the spectral matrix suitable for the master instrument sample after conversion, the symbol "t" representing the conversion, the symbol "M" representing the master instrument; xTRepresenting spectral data measured based on a target instrument, the symbol "T" representing the target instrument;
the spectral correction of the prediction set sample of the target instrument is, according to equation (7):
Xt Mv＝XTvFTtVMt’； (8)
by equation (8), the predicted set spectrum X of the target instrumentTvCorrected to a prediction set spectrum X suitable for the requirements of the host instrumentt MvAnd the symbol "v" denotes a prediction set.
For the sample to be measured by the target instrument, conversion is performed according to the formula (9):
Xt Mu＝XTuFTtVMt’； (9)
the symbol "u" denotes the sample to be measured, XTuRepresenting the spectrum, X, of the sample to be measured by the target instrumentt MuIs XTuThe converted spectrum is suitable for the spectrum of the sample to be detected of the main instrument.
Optionally, n in the step (1) is greater than or equal to 1; the main instrument and the target instrument belong to the same spectrum spectrometer.
The spectrum comprises: infrared, ultraviolet-visible, raman, or nuclear magnetic resonance spectroscopy; the infrared spectrum comprises a mid-infrared spectrum and a near-infrared spectrum;
optionally, the step (2) has the beneficial effects that: for the main instrument and the target instrument, the preprocessing mode of the sample spectrum is kept consistent, so that the interference of non-target factors can be eliminated, and the spectrum profiles obtained by the main instrument and the target instrument are similar.
Optionally, the pretreatment mode of step (2) includes: any one or more of smoothing, first derivative calculation, second derivative calculation, normalization, baseline shift, normal variance, multiple scattering correction, and detrending;
optionally, the number of correction sets in the step (3) is greater than or equal to the number of prediction sets;
optionally, in the step (3), the calibration samples are divided into correction samples and prediction samples, and the dividing manner includes: any one of a KS method, a Rank-KS method, an SPXY method, a Rank-SPXY method and a content gradient method.
Optionally, the number of samples of the transfer set in the step (4) is not too small, otherwise, the contained sample information is not sufficient; if the number of samples of the transfer set is too large, the model transfer effect is not ideal;
optionally, the number of the transfer sets in the step (4) is greater than or equal to 10 and less than or equal to the number of the correction sets;
optionally, the selection criterion of the transfer set in step (4) is to select a plurality of samples from the correction set of a certain index of the main instrument according to the principle that the predicted root mean square error RMSEP is minimum, so that the RMSEP of the same index of the target instrument is minimum, and at this time, the corresponding sample is the transfer set sample; the serial numbers of the transfer set samples are consistent for the master instrument and the target instrument.
Optionally, the spectrum X in the main instrument transfer set sampleMThe spectral variation and wavelength range of (A) are adjusted to match the spectrum X in the target instrument transfer set sampleTThe specific steps of the method for making the spectrum variable consistent with the wavelength range are as follows:
spectrum X collected by main instrumentMIs adjusted to the target instrument spectrum XTThe spectral variation of (d);
calculating the spectrum X collected by the main instrumentMWavelength range of and spectrum X collected by the target instrumentTThe intersection of the wavelength ranges of (a); spectra X collected for primary instruments in the intersection of wavelength rangesMAnd the spectrum X collected by the target instrumentTReserving the spectrum X collected by the main instrument outside the intersection of the wavelength rangesMAnd the spectrum X collected by the target instrumentTAnd deleted.
The calculation method of the conversion matrix in the formulas (2) and (6) comprises the following steps: principal component regression, partial least squares regression.
In the equations (3) and (7), the number of samples of the spectra of the master instrument and the target instrument must be equal, but the number of the variables of the spectra may be equal or may not be equal.
Optionally, in the step (4), principal component decomposition is performed on the target instrument or the sample spectrum data of the transfer set of the main instrument, and the principal component number is determined by combining the target instrument with the capability of interpreting variance of the principal componentThe target RMSEP value is determined by the minimum value, and the maximum value is XTtOr XMtIs determined.
The invention has the advantages of
1. The method can realize the bidirectional transfer of the model, and whether the spectral variables of the main instrument and the target instrument are consistent or not is not limited, and the spectra with different resolutions can be converted mutually, so that the model transfer between spectrometers with different resolutions is realized, and the method is more flexible and has wider application range.
2. The method establishes the internal relation of the spectrums before and after the model transfer through the internal structure of the data, thereby having the effect of inhibiting the influence of noise, better retaining the characteristic information of the original data and improving the success rate of the model transfer.
Drawings
The accompanying drawings, which are incorporated in and constitute a part of this application, illustrate embodiments of the application and, together with the description, serve to explain the application and are not intended to limit the application.
FIG. 1 is a raw near infrared spectrum of a host instrument of an embodiment of the present application;
FIG. 2 is a raw near infrared spectrum of a target instrument of an embodiment of the present application;
FIG. 3 is a near infrared spectrum of a target instrument calibration set sample according to an embodiment of the present application;
FIG. 4 is a near-infrared spectrum of a calibration set sample suitable for a target instrument after the near-infrared spectrum of the calibration set sample of the main instrument is corrected by the model transfer method of the present invention according to an embodiment of the present invention;
FIG. 5 is a graph of the difference between the near infrared spectra of the master instrument calibration set sample and the near infrared spectra of the target instrument calibration set sample after the model transfer according to the present invention;
FIG. 6 is a graph illustrating a sample spectrum of a target instrument prediction set is shifted to a near infrared spectrum suitable for a master instrument prediction set by a PDS method according to an embodiment of the present disclosure;
FIG. 7 shows an example of a method of the present invention for transforming the NIR of a target device prediction set sample to a near IR suitable for a primary device prediction set.
Detailed Description
It should be noted that the following detailed description is exemplary and is intended to provide further explanation of the disclosure. Unless defined otherwise, all technical and scientific terms used herein have the same meaning as commonly understood by one of ordinary skill in the art to which this application belongs.
It is noted that the terminology used herein is for the purpose of describing particular embodiments only and is not intended to be limiting of example embodiments according to the present application. As used herein, the singular forms "a", "an" and "the" are intended to include the plural forms as well, and it should be understood that when the terms "comprises" and/or "comprising" are used in this specification, they specify the presence of stated features, steps, operations, devices, components, and/or combinations thereof, unless the context clearly indicates otherwise.
Example 1: compared with a near infrared model transfer method CN105842190A based on spectral regression in the invention patent of China;
1. theoretical analysis:
the general model transfer method uses high-quality (high-resolution, high-precision) spectrum to correct target spectrum, and according to the principle described in patent (CN105842190A), the near infrared spectrum X of host computermAnd from the near infrared spectrum X of the machinesIs represented by Z each corresponding to a low dimensionmAnd ZsUsing transformation matrices F1Are associated together. Wherein Zm＝Am TXm，AmIs the projection vector set of the host, from which it can be seen that ZmIs XmObtained through certain transformation. Under h characteristic vectors, the projection vector is solved by utilizing a normalized least square method, and then A is obtainedmThus, ZmIs to X under h eigenvectorsmBy approximation of (i.e. passing through) Zm＝Am TXmAfter, ZmRelative to XmThere is a certain loss of information.
In the present invention, however, we do not make any transformation of the main instrument spectrum (see near-infrared of the transfer set samples)Spectrum XMtAnd near infrared spectrum X of the host instrument sampleM) Therefore, the information of the main instrument spectrum is completely preserved without loss. Therefore, theoretically, compared with the method described in the patent (CN105842190A), the method of the present invention can retain the spectral information of the main instrument sample more completely and with less loss during the model transfer process.
2. Comparison of atlas differences before and after model transfer
FIG. 1 is the raw near-infrared spectrum of the host instrument and FIG. 2 shows the raw near-infrared spectrum of the target instrument.
As can be seen from fig. 1 and 2, for the same sample, the difference between the original near infrared spectra of the sample measured by the main instrument and the original near infrared spectra of the sample measured by the target instrument are very large, and the differences are mainly expressed in three aspects of the abscissa representing method and range, the absorbance and the variable number of data. The main instrument and the target instrument respectively represent the abscissa by the wave number and the wavelength, and if converted into the wavelength, the wavelength range measured by the main instrument is 2500nm, 1000-. If the model transfer is carried out, the wavelength ranges need to be unified firstly; the spectral absorbance range measured by the main instrument is 0.39-1.04, and the spectral absorbance range of the target instrument is 0.09-0.34; the number of spectral variables measured by the master instrument is 1557, while the number of spectral variables measured by the target instrument is 125. It can be seen that the model established by the main instrument cannot be used on the target instrument without model transfer.
Fig. 3 shows the near infrared spectrum of the target instrument calibration set sample, and fig. 4 shows the near infrared spectrum of the main instrument calibration set sample after model transfer, which is suitable for the target instrument. These spectra were subjected to a spectral pre-processing (Savitzky-Golay Filtering method, window size 15, polynomial order 2) before model transfer. Fig. 5 shows the difference between the spectra shown in fig. 3 and fig. 4 obtained by the method of the present invention.
As can be seen from FIGS. 3 and 4, even though the difference between the original NIR spectra of the main instrument and the target instrument is large, the NIR spectra of the calibration set of the main instrument are very similar to the NIR spectra of the target instrument after the transfer of the model of the invention. The difference of the maps is very small(see FIG. 5) on the order of 10-6. The difference in the spectra obtained by the method described in the comparative patent (CN105842190A) is of the order of 10-2(see patent CN105842190A in FIG. 3 c). The above results show that: the model transfer method provided by the invention can greatly reduce the spectrum difference between the main instrument and the target instrument, and has very obvious effect.
The unknown sample is predicted based on the near infrared spectrum of the prediction set sample measured by the target instrument, so that the near infrared spectrum suitable for the prediction set sample of the target instrument is obtained without performing model transfer on the near infrared spectrum of the prediction set sample of the main instrument.
Example 2 model Forward transfer
Taking the determination of astragaloside content in the commercial radix astragali extract as an example. The content of astragaloside IV is determined by HPLC liquid chromatography, and the method is described in Chinese pharmacopoeia (2015 edition).
I. Index determination and spectrum collection of samples
A total of 82 samples of commercially available Astragalus membranaceus extract were collected, the astragaloside content was determined by reference to the pharmacopoeia method (2015 edition), and the near infrared spectrum, X, was measured for each of these samples using a Master (Thermo Fisher Scientific, USA) and target (Viavi 1700, USA) instrumentMIndicating the near infrared spectrum, X, measured by the host instrumentTRepresenting the near infrared spectrum measured by the target instrument. The near infrared spectra measured by the master and target instruments are shown in fig. 1 and 2, respectively.
II. Pretreatment of spectra
The pre-processing of the sample spectra remains consistent for both the master and target instruments. A Savitzky-Golay filtration method is used as a spectrum preprocessing method to smooth near infrared spectrum signals of a main instrument and a target instrument, the window size is 15, and the polynomial order is 2.
III, partitioning of sample set
Dividing a sample set into a correction set and a prediction set by adopting a common KS method based on the spectral data of the sample; the number of correction set samples is 42 and the prediction set samples is 40. The serial numbers of the correction set samples and the prediction set samples of the main instrument and the target instrument are consistent.
IV, model transfer and spectral correction
Selecting a certain number of samples as transfer set samples in the correction set samples to construct a transfer matrix so as to realize model transfer; the number of samples of the transfer set is not too small, otherwise, the contained sample information is not sufficient; if the number of samples of the transfer set is too large, the model transfer effect is not ideal; the transfer set samples come from the correction set samples, and the number of the transfer set samples is more than or equal to 10 and less than or equal to the number of the correction set samples; the transfer set samples are determined according to the principle that the RMSEP (root mean square error) is the minimum, a certain number of samples are selected from the correction set samples to enable the RMSEP of a certain index of a target instrument to be the minimum, and the corresponding samples are the transfer set samples at the moment; the serial numbers of the transfer set samples should be consistent for the master and target instruments.
Because the difference between the near infrared spectra of the samples of the main instrument and the target instrument is large, the spectrum variables and the ranges need to be unified before model transfer. As shown in FIGS. 1 and 2, the spectral variation of the master instrument is the wavenumber, which is in the range of 4000cm-1～10000cm-1And the spectral variation of the target instrument is wavelength, ranging from 900nm to 1676 nm. Since the prediction of the unknown sample is based on the spectrum measured by the target instrument, the spectral variables are uniformly wavelength-selected for convenience, and the range selects the intersection of the spectral wavelengths. The spectral variable of the main instrument is converted into wavelength, the wavelength range is 1000nm-2500nm, and then the wavelength range is compared with the range of the target instrument, and the intersection range is 1000 nm-1676 nm.
In the correction set samples, a certain number of transfer set samples are selected to construct a modeling type transfer matrix. Using Kennard-Stone (KS) division method to take different numbers of samples as transfer set samples, and performing principal component decomposition according to the following formula:
XTt＝STtVTt’，
in order to better maintain the spectral information of the target instrument, the principal component number is X according to the interpretable rateTtA certain degree of variance is determined on the basis. In this embodiment, if the number of branch samples is less than 25, the number of principal components is XTtThe rank of (d); if the number of the transferred samples is more than 25, the number of the principal components is set to be 25; will be the X of the main instrumentMtAnd STtPush type STt＝XMtFMtAnd (6) associating. FMtThe calculation of (2) can be solved by adopting a partial least square method, and the corresponding potential factor number is determined by leave-one-out. Different numbers of transfer set samples are tried, and the transfer set samples are preferably determined according to the RMSEP value minimum principle. After transfer set sample determination, FMtAnd is thus determined. By the formula Xt Tc＝XMcFMtVTt' correcting the near infrared spectrum of the main instrument correction set sample to be suitable for the near infrared spectrum of the target instrument correction set. Based on Xt TcAnd establishing a correction model of the target instrument, predicting the astragaloside IV index of a prediction set sample measured by the target instrument, and calculating a predicted root mean square error and a measurement coefficient so as to evaluate the forward model transfer effect.
Table 1 shows the effect evaluation (forward model transmission) of the radix astragali extract before and after the model transfer by the method
As can be seen from Table 1, the spectral resolution measured by the main instrument is high, the accuracy is high, and the model prediction result is optimal. The spectral resolution measured by the target instrument is low, and the model prediction result is the worst (R) before the model is transferredpLowest, RMSEP largest); after the model is transferred by the method, the prediction capability of the model is improved compared with that before the model is transferred, and RpIncreases and RMSEP decreases. This shows that the spectrum modeling performance of the main instrument spectrum, which is suitable for the target instrument and is obtained after model transfer, is improved, and model prediction is enhanced.
Compared with the traditional PDS method, the model prediction result of the method is superior to that of the traditional PDS method, the model transfer effect of the method is better, the model established by the main instrument can be successfully transferred to the target instrument, and the sharing of the model is realized.
Example 3 model reverse transfer
Transferring the main instrument to collect near infrared spectrum data (marked as X)Mt) Through the formula XMt＝SMtVMt' decomposition of principal Components and determination of the appropriate number of principal Components lM. The method of determination was in accordance with example 2 above. The method of selecting the transfer set is also consistent with example 2.
X of target instrumentTtAnd SMtPush type SMt＝XTtFTtAnd (6) associating. Wherein FTtThe calculation of (2) can be solved by adopting a partial least square method, and the corresponding potential factor number is determined by leave-one-out. Different numbers of transfer set samples are tried, and the transfer set samples are preferably determined according to the RMSEP value minimum principle. After transfer set sample determination, FTtAnd is thus determined. Because the unknown sample is predicted by the near infrared spectrum measured by the target instrument, the sample spectrum of the prediction set of the target instrument is predicted by the formula Xt Mv＝XTvFTtVMt' corrected to fit the near infrared spectrum of the master instrument prediction set sample. And predicting the content of astragaloside in the prediction set sample through the established main instrument correction set model, and calculating the prediction root mean square error and the determination coefficient so as to evaluate the transfer effect of the reverse model. A comparison of the maps after the inverse model transfer is shown in fig. 6 and 7.
As can be seen from fig. 6 and 7, the transfer effect of the method is significantly better than that of the conventional PDS method. The PDS method is based on direct correction of the data of the spectral signals measured by the target instrument and the main instrument, so that the spectral signals measured by the target instrument have few data points and low resolution, while the spectral signals measured by the main instrument have many data points and high resolution. When the PDS is used for correction, due to the fact that the spectrum signal data points of the target instrument are few and the information quantity is insufficient, the spectrum suitable for the main instrument after correction has some abnormal phenomena. The method adopts the principle component correlation of target instrument spectrum and main instrument transfer set spectrum (see S)Mt＝XTtFTt) In which S isMtIntegrates a large amount of information of a transfer set sample of a main instrument, and uses SMt＝XTtFTtCorrelating to obtain the light of the target instrument as much as possibleThe spectral information is transmitted, and the calculation result is S under the principle of least squareMtThus, the information of the spectrum of the main instrument transfer set is well preserved. At the same time, using formula Xt Mv＝XTvFTtVMt' when spectrum fitting the prediction set of the main instrument is obtained, VMt' derived from the spectrum of the master instrument transfer set, the matrix also carrying the spectral information of the master instrument transfer set, so Xt MvThe spectral information of the main instrument can be well reserved. As can be seen from FIG. 7, the prediction set spectrum suitable for the main instrument obtained by the model transfer of the method is smooth, and no abnormal phenomenon occurs. The results of the astragalus extract before and after model transfer according to the method (inverse model transfer) are shown in table 2.
TABLE 2 Effect evaluation of Astragalus membranaceus extract before and after model transfer by this method (reverse model transfer)
As can be seen from Table 2, the method has better model transfer effect, the prediction capability is improved compared with that before model transfer, the prediction result is close to the result of the main instrument, and the success rate of model transfer is improved; and the traditional PDS method fails to transfer the model because the map has an abnormal phenomenon.
The above description is only a preferred embodiment of the present application and is not intended to limit the present application, and various modifications and changes may be made by those skilled in the art. Any modification, equivalent replacement, improvement and the like made within the spirit and principle of the present application shall be included in the protection scope of the present application.
Claims (7)
1. A model transfer method based on spectral data is characterized by comprising the following steps:
step (1): index determination and spectrum collection of samples: aiming at a plurality of calibration samples, n indexes are measured by adopting an industry standard detection method; respectively using a main instrument and a target instrument to collect the spectrums of the calibration samples to obtain the collection of the main instrumentSpectrum X of the setMAnd the spectrum X collected by the target instrumentT；
Step (2): pretreatment of the spectrum: spectrum X collected from the host instrumentMAnd the spectrum X collected by the target instrumentTRespectively adopting the same mode to carry out pretreatment;
and (3): dividing the calibration samples into correction samples and prediction samples, and acquiring the spectrum X of the main instrument according to the correction samples and the prediction samplesMDividing the samples into correction set samples and prediction set samples; spectrum X collected by target instrument according to correction sample and prediction sampleTDividing the samples into correction set samples and prediction set samples;
and (4): model transfer is carried out from the main instrument to the target instrument, namely forward transfer:
a step (401): selecting a set number of samples from the main instrument correction set samples as transfer set samples, then selecting samples with the same serial number from the target instrument correction set as target instrument transfer set samples, wherein the transfer set samples of the main instrument correspond to the serial numbers of the target instrument transfer set samples one by one; the number of samples of the spectra of the master and target instruments must be equal, but the number of variables of the spectra may or may not be equal; transferring the spectrum X in the sample set of the main instrumentMThe spectral variation and wavelength range of (A) are adjusted to match the spectrum X in the target instrument transfer set sampleTThe spectral variation of (a) is consistent with the wavelength range;
step (402): correlating the spectral intensity data of the main instrument transfer set sample with the spectral intensity data of the target instrument transfer set sample to obtain a relation model, and obtaining a conversion matrix;
step (403): substituting the spectral data of the main instrument correction set sample into the conversion matrix to obtain the spectral data suitable for the target instrument correction set sample;
a step (404): establishing a target instrument correction model based on the spectral data suitable for the target instrument correction set sample: the spectral data suitable for the target instrument correction set sample are used as independent variables, n measuring indexes are used as dependent variables, a multivariate correction method is adopted to correlate the spectral data suitable for the target instrument correction set sample with the n index data, and n correction models are established;
step (405): collecting a spectrum of a sample to be detected by using a target instrument, and predicting an index value of the sample by using a target instrument correction model;
in the forward transfer process, the target instrument transfer set spectrum XTt is subjected to principal component decomposition to obtain:
XTt＝STtVTt’；(1)
wherein STt represents a principal component score matrix, VTt represents a principal component load matrix, the symbol "'" represents the transpose of the principal component load matrix, the subscript "T" represents the target instrument, and the subscript "T" represents the transfer set;
setting the number lT of main components; correlating the master instrument transfer set spectrum XMt with STt:
STt＝XMtFMt；(2)
wherein FMt denotes a transformation matrix, the symbol "M" denotes a master, and "t" denotes a transition set;
calculating and solving according to FMt ═ X + Mt STt, wherein a sign of "+" represents generalized inverse operation; x + Mt represents the generalized inverse matrix of XMt;
substituting the formula (2) into the formula (1) to obtain the spectrum after model conversion:
XtT＝XMFMtVTt’；(3)
wherein XtT represents the spectral matrix suitable for the target instrument sample after transformation, the symbol "T" represents the target instrument, XM represents the spectral data measured based on the master instrument;
the spectra of the master instrument calibration set are calibrated as follows equation (3):
XtTc＝XMcFMtVTt’；(4)
correcting the spectra XMc of the master instrument calibration set to the spectra XtTc appropriate for the target instrument by equation (4), where the symbol "c" represents the calibration set;
the step (4) is replaced by the following steps: model transfer is carried out from the target instrument to the main instrument, namely reverse transfer:
step (411): selecting a set number of samples from the main instrument correction set samples as transfer set samples, then selecting samples with the same serial number from the target instrument correction set as target instrument transfer set samples, wherein the transfer set samples of the main instrument correspond to the target instrument transfer set samples one by one; adjusting the spectral variable and wavelength range of the spectrum XM in the sample of the transfer set of the main instrument to be consistent with the spectral variable and wavelength range of the spectrum XT in the sample of the transfer set of the target instrument;
step (412): correlating the spectral intensity data of the target instrument transfer set sample with the spectral intensity data of the main instrument transfer set sample to obtain a relation model, and obtaining a conversion matrix;
step (413): collecting the spectrum of the sample to be detected by using the target instrument, and correcting the spectrum of the sample to be detected of the target instrument into the spectrum of the sample to be detected suitable for the main instrument based on the relation model;
a step (414): establishing a correction model by using the spectrum of the main instrument correction set sample: taking spectral data of a correction set sample of a main instrument as an independent variable, taking n measurement indexes as dependent variables, and establishing n correction models by adopting a multivariate correction method; replacing the spectrum of the sample to be predicted which is suitable for the main instrument into a correction model of the main instrument to predict the index value of the sample to be predicted;
in the reverse transfer process, the main component decomposition is carried out on the transfer set spectrum XMt of the main instrument to obtain:
XMt＝SMtVMt’；(5)
wherein SMt represents the score matrix of the principal component, "M" represents the master, and "t" represents the transfer set; VMt denotes the principal component loading matrix, the symbol "'" denotes the transpose of the matrix;
set principal component numbers lM, associate target instrument transfer set spectra XTt with SMt:
SMt＝XTtFTt；(6)
wherein FTt denotes a transformation matrix, the symbol "T" denotes a target instrument, and "T" denotes a transition set;
wherein the solution is calculated according to FTt ═ X + Tt SMt; x + Tt represents a generalized inverse matrix of XTt;
substituting equation (6) into equation (5), model-converted spectra:
XtM＝XTFTtVMt’；(7)
wherein XtM denotes the spectral matrix adapted to the master instrument sample after conversion, the symbol "t" denotes the conversion, and the symbol "M" denotes the master instrument; XT represents spectral data measured based on the target instrument, and symbol "T" represents the target instrument;
the spectral correction of the prediction set sample of the target instrument is, according to equation (7):
XtMv＝XTvFTtVMt’；(8)
by equation (8), the prediction set spectrum XTv of the target instrument is corrected to the prediction set spectrum XtMv that fits the requirements of the master instrument, the symbol "v" representing the prediction set;
for the sample to be measured by the target instrument, conversion is performed according to the formula (9):
XtMu＝XTuFTtVMt’；(9)
the symbol "u" represents the sample to be measured, XTu represents the spectrum of the sample to be measured by the target instrument, and XtMu is the spectrum of the sample to be measured which is XTu transformed to fit the main instrument.
2. A method of model transfer based on spectral data as claimed in claim 1, further comprising after step (404) and before step (405): and substituting the spectrum of the target instrument prediction set into the correction model, calculating the predicted values of the n indexes of each prediction set sample through the correction model, comparing the predicted values with the actual measured values, and performing model evaluation.
3. A method of model transfer based on spectral data as claimed in claim 1, further comprising after step (414): and correcting the spectrum of the prediction set of the target instrument into a prediction sample spectrum suitable for the main instrument by adopting a relational model, substituting the prediction sample spectrum into a correction model, calculating the predicted values of n indexes of the samples of each prediction set through the correction model, comparing the predicted values with the actual measured values, and performing model evaluation.
4. The method for model transfer based on spectral data according to claim 1, wherein n is greater than or equal to 1 in step (1); the main instrument and the target instrument belong to spectrometers of the same spectrum; the spectrum comprises: infrared, ultraviolet-visible, raman, or nuclear magnetic resonance spectroscopy; the infrared spectrum comprises a mid-infrared spectrum and a near-infrared spectrum;
the pretreatment mode of the step (2) comprises the following steps: any one or more of smoothing, first derivative calculation, second derivative calculation, normalization, baseline shift, normal variance, multiple scatter correction, and detrending.
5. The method of claim 1, wherein the number of correction sets in step (3) is greater than or equal to the number of prediction sets; in the step (3), the calibration samples are divided into correction samples and prediction samples, and the division mode includes: any one of a KS method, a Rank-KS method, an SPXY method, a Rank-SPXY method and a content gradient method.
6. The method of claim 1, wherein the model is a model transfer method based on spectral data,
the number of the transfer sets in the step (4) is more than or equal to 10 and less than or equal to the number of the correction sets;
selecting a plurality of samples from a correction set of a certain index of a main instrument according to the principle of predicting the minimum root mean square error RMSEP, so that the RMSEP of the same index of a target instrument is minimum, and the corresponding sample is the sample of the transfer set at the moment; the serial numbers of the transfer set samples are consistent for the master instrument and the target instrument.
7. The method of claim 1, wherein the model is a model transfer method based on spectral data,
spectrum X in the main instrument transfer set sampleMThe spectral variation and wavelength range of (A) are adjusted to match the spectrum X in the target instrument transfer set sampleTThe specific steps of the method for making the spectrum variable consistent with the wavelength range are as follows:
spectrum X collected by main instrumentMIs adjusted to the target instrument spectrum XTThe spectral variation of (d);
calculating main instrumentSpectrum X collected by the deviceMWavelength range of and spectrum X collected by the target instrumentTThe intersection of the wavelength ranges of (a); spectra X collected for primary instruments in the intersection of wavelength rangesMAnd the spectrum X collected by the target instrumentTReserving the spectrum X collected by the main instrument outside the intersection of the wavelength rangesMAnd the spectrum X collected by the target instrumentTAnd deleted.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201811271230.6A CN109444066B (en) | 2018-10-29 | 2018-10-29 | Model transfer method based on spectral data |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201811271230.6A CN109444066B (en) | 2018-10-29 | 2018-10-29 | Model transfer method based on spectral data |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN109444066A CN109444066A (en) | 2019-03-08 |
| CN109444066B true CN109444066B (en) | 2020-04-14 |
Family
ID=65549356
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201811271230.6A Active CN109444066B (en) | 2018-10-29 | 2018-10-29 | Model transfer method based on spectral data |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN109444066B (en) |
Families Citing this family (9)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN108509997A (en) * | 2018-04-03 | 2018-09-07 | 深圳市药品检验研究院（深圳市医疗器械检测中心） | A method of Chemical Pattern Recognition is carried out to the true and false that Chinese medicine Chinese honey locust is pierced based on near-infrared spectrum technique |
| CN111198165A (en) * | 2020-01-14 | 2020-05-26 | 重庆理工大学 | Method for measuring water quality parameters based on spectral data standardization |
| CN111337452A (en) * | 2020-04-08 | 2020-06-26 | 四川虹微技术有限公司 | Method for verifying feasibility of spectral data model transfer algorithm |
| CN111563436B (en) * | 2020-04-28 | 2022-04-08 | 东北大学秦皇岛分校 | Infrared spectrum measuring instrument calibration migration method based on CT-CDD |
| CN111579526B (en) * | 2020-06-08 | 2023-05-30 | 贵州中烟工业有限责任公司 | Method for representing difference and correction of near infrared instrument |
| CN112461770B (en) * | 2020-11-17 | 2022-11-29 | 山东省科学院海洋仪器仪表研究所 | Method for acquiring performance of spectrometer |
| CN112617824B (en) * | 2020-12-09 | 2021-10-01 | 南京睿蜂健康医疗科技有限公司 | Ice drug addict detection system based on multichannel fNIRS signals |
| CN112683816B (en) * | 2020-12-25 | 2021-08-06 | 中船重工安谱（湖北）仪器有限公司 | Spectrum identification method for spectrum model transmission |
| CN113762208B (en) * | 2021-09-22 | 2023-07-28 | 山东大学 | Spectrum conversion method of near infrared spectrum and characteristic spectrum and application thereof |
Citations (6)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN101832922A (en) * | 2010-05-19 | 2010-09-15 | 中国农业大学 | Method for transferring near infrared model of organic fertilizer product |
| CN103018178A (en) * | 2012-12-06 | 2013-04-03 | 江苏省质量安全工程研究院 | LPP (Local Preserving Projection)-based Infrared spectrometer calibration method |
| CN104089911A (en) * | 2014-06-27 | 2014-10-08 | 桂林电子科技大学 | Spectral model transmission method based on unary linear regression |
| CN105738317A (en) * | 2016-02-16 | 2016-07-06 | 广州纤维产品检测研究院 | Textile near-infrared model transfer method |
| CN105842190A (en) * | 2016-03-17 | 2016-08-10 | 浙江中烟工业有限责任公司 | Near-infrared model transfer method based on spectral regression |
| CN106845548A (en) * | 2017-01-24 | 2017-06-13 | 清华大学 | Model transfer method based on high-spectral data |
-
2018
- 2018-10-29 CN CN201811271230.6A patent/CN109444066B/en active Active
Patent Citations (6)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN101832922A (en) * | 2010-05-19 | 2010-09-15 | 中国农业大学 | Method for transferring near infrared model of organic fertilizer product |
| CN103018178A (en) * | 2012-12-06 | 2013-04-03 | 江苏省质量安全工程研究院 | LPP (Local Preserving Projection)-based Infrared spectrometer calibration method |
| CN104089911A (en) * | 2014-06-27 | 2014-10-08 | 桂林电子科技大学 | Spectral model transmission method based on unary linear regression |
| CN105738317A (en) * | 2016-02-16 | 2016-07-06 | 广州纤维产品检测研究院 | Textile near-infrared model transfer method |
| CN105842190A (en) * | 2016-03-17 | 2016-08-10 | 浙江中烟工业有限责任公司 | Near-infrared model transfer method based on spectral regression |
| CN106845548A (en) * | 2017-01-24 | 2017-06-13 | 清华大学 | Model transfer method based on high-spectral data |
Non-Patent Citations (1)
| Title |
|---|
| 基于拓展光谱的近红外光谱模型转移方法;王家俊 等;《中国烟草学报》;20141231;第20卷(第6期);1-5、11 * |
Also Published As
| Publication number | Publication date |
|---|---|
| CN109444066A (en) | 2019-03-08 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN109444066B (en) | Model transfer method based on spectral data | |
| Andersson | Direct orthogonalization | |
| Fan et al. | Calibration model transfer for near-infrared spectra based on canonical correlation analysis | |
| Ye et al. | Non-destructive prediction of protein content in wheat using NIRS | |
| Huang et al. | Improved generalization of spectral models associated with Vis-NIR spectroscopy for determining the moisture content of different tea leaves | |
| CN111563436B (en) | Infrared spectrum measuring instrument calibration migration method based on CT-CDD | |
| CN104089911A (en) | Spectral model transmission method based on unary linear regression | |
| CN101413884B (en) | Near-infrared spectrum analyzer and method for correcting resolution | |
| Zhang et al. | The method of calibration model transfer by optimizing wavelength combinations based on consistent and stable spectral signals | |
| EP3265780B1 (en) | Optimized spectral matching and display | |
| CN104990895B (en) | A kind of near infrared spectrum signal standards normal state bearing calibration based on regional area | |
| WO2023046216A1 (en) | Graph conversion method for near infrared spectrum and feature graph, and use thereof | |
| CN109507315A (en) | Complex samples GC-MS, which is automatically parsed, realizes that compound accurately identifies and the method for otherness component screening | |
| Yun-Ying et al. | Progress and applications of multivariate calibration model transfer methods | |
| CN105628645A (en) | Infrared analysis model transfer method | |
| Fan et al. | Direct calibration transfer to principal components via canonical correlation analysis | |
| Shen et al. | Rapid and real-time detection of moisture in black tea during withering using micro-near-infrared spectroscopy | |
| Oliveri et al. | A spectral transfer procedure for application of a single class-model to spectra recorded by different near-infrared spectrometers for authentication of olives in brine | |
| Wang et al. | A dual model strategy to transfer multivariate calibration models for near-infrared spectral analysis | |
| Boysworth et al. | Aspects of multivariate calibration applied to near-infrared spectroscopy | |
| Nørgaard | A multivariate chemometric approach to fluorescence spectroscopy | |
| Zeng et al. | Mixture analysis using non‐negative elastic net for Raman spectroscopy | |
| CN111157484A (en) | Near infrared spectrum model transfer method for fruit sugar degree detection equipment | |
| Huang et al. | Multigranularity cascade forest algorithm based on hyperspectral imaging to detect moisture content in Daqu | |
| CN112683816B (en) | Spectrum identification method for spectrum model transmission |
Legal Events
| Date | Code | Title | Description |
|---|---|---|---|
| PB01 | Publication | ||
| PB01 | Publication | ||
| SE01 | Entry into force of request for substantive examination | ||
| SE01 | Entry into force of request for substantive examination | ||
| GR01 | Patent grant | ||
| GR01 | Patent grant |