WO2023046216A1 - 一种近红外光谱与特征图谱的图谱转换方法及其应用 - Google Patents
一种近红外光谱与特征图谱的图谱转换方法及其应用 Download PDFInfo
- Publication number
- WO2023046216A1 WO2023046216A1 PCT/CN2022/132598 CN2022132598W WO2023046216A1 WO 2023046216 A1 WO2023046216 A1 WO 2023046216A1 CN 2022132598 W CN2022132598 W CN 2022132598W WO 2023046216 A1 WO2023046216 A1 WO 2023046216A1
- Authority
- WO
- WIPO (PCT)
- Prior art keywords
- spectrum
- matrix
- characteristic
- instrument
- samples
- Prior art date
Links
- 238000000034 method Methods 0.000 title claims abstract description 92
- 238000006243 chemical reaction Methods 0.000 title claims abstract description 68
- 238000002329 infrared spectrum Methods 0.000 title claims abstract description 41
- 238000001228 spectrum Methods 0.000 claims abstract description 151
- 239000011159 matrix material Substances 0.000 claims abstract description 78
- 238000007781 pre-processing Methods 0.000 claims abstract description 19
- 230000002159 abnormal effect Effects 0.000 claims abstract description 14
- 238000000354 decomposition reaction Methods 0.000 claims abstract description 10
- 230000008030 elimination Effects 0.000 claims abstract description 6
- 238000003379 elimination reaction Methods 0.000 claims abstract description 6
- 230000003595 spectral effect Effects 0.000 claims description 16
- 238000012937 correction Methods 0.000 claims description 14
- 239000003814 drug Substances 0.000 claims description 14
- 238000012545 processing Methods 0.000 claims description 9
- 238000004364 calculation method Methods 0.000 claims description 6
- 238000012360 testing method Methods 0.000 claims description 6
- 238000001514 detection method Methods 0.000 claims description 4
- 238000005259 measurement Methods 0.000 claims description 3
- 238000005457 optimization Methods 0.000 claims description 3
- 238000004164 analytical calibration Methods 0.000 claims description 2
- 238000009499 grossing Methods 0.000 claims description 2
- 238000010606 normalization Methods 0.000 claims description 2
- 238000013450 outlier detection Methods 0.000 claims description 2
- 230000008569 process Effects 0.000 claims description 2
- 230000001131 transforming effect Effects 0.000 claims 1
- 238000004587 chromatography analysis Methods 0.000 abstract description 4
- 230000017105 transposition Effects 0.000 abstract 1
- 239000000523 sample Substances 0.000 description 36
- 238000004458 analytical method Methods 0.000 description 14
- 230000000694 effects Effects 0.000 description 9
- 238000004497 NIR spectroscopy Methods 0.000 description 7
- 238000010586 diagram Methods 0.000 description 6
- 238000005516 engineering process Methods 0.000 description 6
- 230000001066 destructive effect Effects 0.000 description 5
- 238000012546 transfer Methods 0.000 description 5
- 239000007788 liquid Substances 0.000 description 4
- 239000000463 material Substances 0.000 description 4
- 239000010365 xinkeshu Substances 0.000 description 4
- 238000011160 research Methods 0.000 description 3
- UFHFLCQGNIYNRP-UHFFFAOYSA-N Hydrogen Chemical compound [H][H] UFHFLCQGNIYNRP-UHFFFAOYSA-N 0.000 description 2
- 239000008186 active pharmaceutical agent Substances 0.000 description 2
- 238000012512 characterization method Methods 0.000 description 2
- 230000000052 comparative effect Effects 0.000 description 2
- 238000011161 development Methods 0.000 description 2
- 229910052739 hydrogen Inorganic materials 0.000 description 2
- 239000001257 hydrogen Substances 0.000 description 2
- 238000007689 inspection Methods 0.000 description 2
- 238000004519 manufacturing process Methods 0.000 description 2
- 238000012986 modification Methods 0.000 description 2
- 230000004048 modification Effects 0.000 description 2
- 238000000513 principal component analysis Methods 0.000 description 2
- 238000011057 process analytical technology Methods 0.000 description 2
- 230000009466 transformation Effects 0.000 description 2
- 238000012795 verification Methods 0.000 description 2
- 230000003044 adaptive effect Effects 0.000 description 1
- 230000009286 beneficial effect Effects 0.000 description 1
- 230000007812 deficiency Effects 0.000 description 1
- 238000002474 experimental method Methods 0.000 description 1
- 238000004128 high performance liquid chromatography Methods 0.000 description 1
- 239000003348 petrochemical agent Substances 0.000 description 1
- 238000004886 process control Methods 0.000 description 1
- 239000000047 product Substances 0.000 description 1
- 238000007430 reference method Methods 0.000 description 1
- 230000004044 response Effects 0.000 description 1
- 238000000926 separation method Methods 0.000 description 1
- 238000004611 spectroscopical analysis Methods 0.000 description 1
- 239000004575 stone Substances 0.000 description 1
- 230000004304 visual acuity Effects 0.000 description 1
Images
Classifications
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F2218/00—Aspects of pattern recognition specially adapted for signal processing
- G06F2218/08—Feature extraction
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F17/00—Digital computing or data processing equipment or methods, specially adapted for specific functions
- G06F17/10—Complex mathematical operations
- G06F17/16—Matrix or vector computation, e.g. matrix-matrix or matrix-vector multiplication, matrix factorization
Definitions
- the invention belongs to the technical field of chromatographic analysis, and relates to a method for converting near-infrared spectra and characteristic maps and an application thereof.
- NIR Near-infrared spectroscopy
- GAC green analytical chemistry
- Near-infrared spectroscopy is fast, non-destructive, and green, and is suitable for on-site inspection and real-time online analysis. It has become an important part of process analytical technology (PAT), but the peaks of near-infrared spectroscopy overlap seriously, and the specific component information is not clear.
- PAT process analytical technology
- Fingerprint refers to the spectrum or image that can characterize or reflect the characteristics of a certain Chinese medicinal material or Chinese patent medicine obtained by using modern analysis methods after proper processing.
- the application of characteristic map (or fingerprint) map technology in the quality standard of traditional Chinese medicine has improved the research ideas, methods and models of the quality standard of the complex system of traditional Chinese medicine, and is also the development trend of the overall controllability of the quality of traditional Chinese medicine in the future.
- High-performance liquid chromatography (HPLC) has high separation efficiency, strong characteristics, and good reliability, and is suitable as a reference method for verification.
- HPLC high-performance liquid chromatography
- its pretreatment is complicated, it is difficult to be used for direct analysis of samples, and it has the potential to destroy the original shape of samples, take a long time, and pollute the environment. Therefore, chromatographic analysis technology is difficult to apply to the timely and rapid analysis of the production process of traditional Chinese medicine and its intermediate materials, and it is difficult to do large-scale inspections. It is usually used for small sample detection and verification after the production of intermediate materials
- SBC univariate slope-intercept correction
- SST spectral space transformation
- DS direct correction
- PDS piecewise correction
- NIR near-infrared
- spectroscopy analysis technology reflects all the information of hydrogen-containing groups
- the material components of traditional Chinese medicine can be determined by fingerprints, or by near-infrared spectroscopy. to test.
- the fast near-infrared (NIR) spectrometry method is associated with the strong resolving power and reliable feature (or fingerprint) map method based on chromatography technology, the conversion from near-infrared spectrum to feature (or fingerprint) map can be achieved.
- the NIR spectrum is fast, non-destructive, and green, and the converted characteristic (or fingerprint) spectrum also has the characteristics of strong resolution and good reliability, which can give full play to the advantages of the two analysis techniques.
- the present invention provides a method for converting near-infrared spectrum and characteristic spectrum and its application, the method can realize the conversion between near-infrared spectrum and characteristic (or fingerprint) spectrum, and the two There is no limit to whether the number of spectrum variables of the instrument is consistent, and it is possible to establish a connection between the spectrum and the characteristic (or fingerprint) spectrum, and realize the conversion of spectra between different types of instruments.
- T means transpose
- the conversion direction is to convert the near-infrared spectrum of the sample to the corresponding chromatographic feature (or fingerprint) spectrum. According to the algorithm described later, the reverse, that is, the conversion from the chromatographic feature (or fingerprint) spectrum to the NIR spectrum is also theoretically feasible.
- the method can realize the spectrum conversion between the near-infrared spectrum and the feature (or fingerprint) spectrum (theoretically can be converted), and there is no limit to whether the number of spectrum variables of the two is consistent, and the spectrum and the feature ( or fingerprint) spectra to establish a link to realize the transfer of spectra between different types of instruments, so the method is more flexible and has a wider scope of application.
- the method combines the strong characterizing ability of the component characteristics of the chromatographic characteristic (or fingerprint) spectrum of the traditional Chinese medicine with the fast and non-destructive characteristics of the near-infrared spectrum, and fully exerts the advantages of the two analysis techniques.
- the method can better retain the characteristic information of the original data of the converted map.
- Fig. 1 the original near-infrared spectrum of all samples in embodiment 1;
- Fig. 2 the original characteristic (or fingerprint) spectrum of all samples in embodiment 1;
- Fig. 3 the principal component projection diagram of removing the abnormal sample in the near-infrared spectrum in embodiment 1;
- Fig. 4 Principal component projection diagram of abnormal samples removed from feature (or fingerprint) spectrum in embodiment 1;
- Fig. 5 embodiment 1 from the comparative atlas before and after the atlas conversion of the present invention from the instrument prediction set;
- Fig. 6 the characteristic peak that embodiment 1 selects from instrument characteristic (or fingerprint) collection of spectra
- Fig. 8 the raw feature (or fingerprint) spectrum of all samples in embodiment 2;
- Fig. 9 The principal component projection diagram of the abnormal sample removed from the near-infrared spectrum in Example 2;
- Fig. 10 Principal component projection diagram of abnormal samples removed from feature (or fingerprint) spectrum in embodiment 2;
- Fig. 11 Example 2 from the comparative atlas before and after the atlas conversion of the present invention from the instrument prediction set;
- Fig. 12 The characteristic peaks selected from the instrument characteristic (or fingerprint) spectrum in Example 2.
- the present invention provides a conversion method of near infrared spectrum and characteristic spectrum and its application.
- the conversion direction is from the near-infrared spectrum of the sample to the corresponding chromatographic characteristic spectrum, and it can also be reversed, that is, from the chromatographic characteristic spectrum to the near-infrared spectrum, which specifically includes the following steps:
- Spectral preprocessing preprocessing the original spectral matrix X 1 after removing outliers, and performing singular value decomposition to obtain the score matrix S 1 ;
- Preprocessing of the characteristic map preprocessing the original matrix X 2 of the characteristic map collected from the instrument after removing outliers
- Division of sample sets divide several calibration samples after preprocessing into calibration samples and prediction samples, and divide the original spectrum matrix X of the main instrument into calibration set samples X 1mod and prediction set samples X 1test ;
- the original matrix X2 of the characteristic map of the instrument is divided into a calibration set sample X 2trans_mod and a prediction set sample X 2trans_test ;
- the calibration samples of the master instrument and the slave instrument are in one-to-one correspondence with the prediction samples;
- T s and P s are the score and loading matrix of the combined S comb matrix, respectively; E represents the corresponding error part; the superscript "T” represents the transpose; the subscripts "s” and “n” represent the spectral information and Noise response factor.
- the superscript "+" indicates the pseudo-inverse of the matrix.
- X 1 can be used as the feature (or fingerprint) spectrum
- X 2 can be used as the NIR spectrum
- step (2) abnormal values are detected by the Hotelling T 2 method.
- the preprocessing method of the spectrum includes: smoothing, first derivative calculation, second derivative calculation, normalization processing, baseline drift processing, standard normal variable processing, multiple scatter correction It is also possible to correct the spectrum without using preprocessing methods.
- the preprocessing method for the feature (or fingerprint) spectrum includes any one of the correlation optimization warping method and the adaptive iterative weighted least squares method.
- the number of the correction set is greater than or equal to the number of the prediction set, and the setting ratio of the number of samples in the correction set and the prediction set is 2:1 or above.
- step (5) the several calibration samples are divided into calibration samples and prediction samples, and the division methods include: KS method, Rank-KS method, SPXY method, Rank-SPXY method and content gradient method. A sort of.
- step (6) the number of samples of the near-infrared spectrum and characteristic (or fingerprint) spectrum of the method must be the same, and the measured samples will correspond one-to-one, but the variables of the spectrum and the characteristic (or fingerprint) spectrum of the spectrum Numbers can be equal or unequal, and the scope and content of application are wider.
- Adopt method of the present invention to be able to set up matrix conversion relation (as formula (8), (9) and (10)) with NIR spectrum and characteristic (or fingerprint) collection of illustrative plates, realize between near-infrared spectrum and characteristic (or fingerprint) collection of illustrative plates Spectrum conversion, and there is no limit to whether the number of spectral variables of the two is consistent, can establish a connection between the spectrum and the characteristic (or fingerprint) spectrum, and realize the conversion of the spectrum between different types of instruments, so the method of the present invention is more flexible and applicable. more extensive.
- the method of the invention combines the characteristics of strong characterization ability of the component characteristics of the chromatographic characteristic (or fingerprint) spectrum of the traditional Chinese medicine with the characteristics of rapidity and non-destructiveness of the near-infrared spectrum, and fully exerts the advantages of the two analysis techniques.
- X 1 is the near-infrared spectrum matrix of the sample, which was measured by the Fourier transform near-infrared spectrometer (Antaris II, Thermo Fisher, USA) as the main instrument.
- the original spectrum of the sample is shown in Figure 1.
- X2 is the feature (or fingerprint) spectrum matrix of the sample, which is measured by a high-performance liquid chromatograph (Agilent 1260, Agilent Technologies Co., Ltd., USA) as a secondary instrument.
- the original feature (or fingerprint) spectrum of the sample is shown in Figure 2.
- the sample spectrum in this implementation example does not use the preprocessing method for spectral processing.
- the sample characteristic (or fingerprint) spectrum is corrected by the correlation optimization warping (COW) method, and the purpose is to align the characteristic peaks of the characteristic (or fingerprint) spectrum.
- COW correlation optimization warping
- the commonly used KS method was used to divide the sample set into a calibration set and a prediction set; the number of samples in the calibration set was 12, and the number of samples in the prediction set was 5. The serial numbers of the calibration set samples and prediction set samples of the master instrument and the slave instrument should be consistent.
- the converted spectrum can be corrected to be suitable for the characteristic (or fingerprint) spectrum of the slave instrument, the slave instrument before and after conversion
- the calibration set spectrum is shown in Figure 5, and the converted samples from the instrument prediction set are predicted, and the relative analysis error (RPD), prediction root mean square error (RMSEP) and similarity (Similarity ), to evaluate the map conversion effect.
- RPD relative analysis error
- RMSEP prediction root mean square error
- similarity similarity
- X 1 is the near-infrared spectrum matrix of the sample, which is measured by the Fourier transform near-infrared spectrometer (Antaris II, Thermo Fisher, USA) as the main instrument.
- the original spectrum of the sample is shown in Figure 7.
- X2 is the characteristic (or fingerprint) spectrum matrix of the sample, which is measured by a high performance liquid chromatograph (Agilent 1260, Agilent Technologies Co., Ltd., USA) as a secondary instrument.
- the original characteristic (or fingerprint) spectrum of the sample is shown in Figure 8.
- the sample spectra in this example are preprocessed using the Standard Normal Variation (SNV) method.
- SNV Standard Normal Variation
- the sample characteristic (or fingerprint) spectrum is corrected by the COW method, and the purpose is to align the characteristic peaks of the characteristic (or fingerprint) spectrum.
- the commonly used Kennard-Stone (KS) method was used to divide the sample set into a calibration set and a prediction set; the number of samples in the calibration set was 14, and the number of samples in the prediction set was 10. The serial numbers of the calibration set samples and prediction set samples of the master instrument and the slave instrument should be consistent.
- the converted spectrum can be corrected to be suitable for the characteristic (or fingerprint) spectrum of the slave instrument, and the slave instrument before and after conversion
- the calibration set spectrum is shown in Figure 11, and the converted samples from the instrument prediction set are predicted, and the relative analysis error (RPD), prediction root mean square error (RMSEP) and similarity (Similarity ), to evaluate the map conversion effect.
- RPD relative analysis error
- RMSEP prediction root mean square error
- similarity similarity
- the ratio of the peak areas of the characteristic peaks before and after conversion is basically between 0.75 and 1.25, and the average value is 1.02, which is close to 1. It can be considered that before and after conversion The result is acceptable, and it further illustrates that the Xinkeshu tablet solution can have a good spectrum conversion effect by the method of the present invention.
- the spectrum obtained after the conversion of the near-infrared spectrum of the master instrument through the spectrum conversion method described in the present invention is different from the characteristic (or fingerprint) spectrum of the instrument.
- the map conversion method is not limited to the same number of map variables, and has wider scope and content.
Landscapes
- Engineering & Computer Science (AREA)
- Physics & Mathematics (AREA)
- Theoretical Computer Science (AREA)
- General Physics & Mathematics (AREA)
- Mathematical Physics (AREA)
- Pure & Applied Mathematics (AREA)
- General Engineering & Computer Science (AREA)
- Mathematical Analysis (AREA)
- Computer Vision & Pattern Recognition (AREA)
- Mathematical Optimization (AREA)
- Data Mining & Analysis (AREA)
- Computational Mathematics (AREA)
- Algebra (AREA)
- Software Systems (AREA)
- Databases & Information Systems (AREA)
- Computing Systems (AREA)
- Artificial Intelligence (AREA)
- Signal Processing (AREA)
- Investigating Or Analysing Materials By Optical Means (AREA)
Abstract
本发明属于色谱分析技术领域,涉及一种近红外光谱与特征图谱的图谱转换方法及其应用,包括以下步骤:首先,将光谱原始矩阵X 1和特征图谱原始矩阵X 2进行异常值剔除及预处理,然后进行奇异值分解,在保留相同的主成分数下,得到X 1的得分矩阵S 1和X 2的S 2;将S 1和S 2两个矩阵进行关联;通过公式X 2trans=[X 1V 1(P 1 T) +(P 2 T)]V 2 T,将转换后的图谱校正为适合从仪器的特征图谱;其中,X 2trans表示转移后的特征图谱矩阵;V 1的含义是X 1的负载矩阵;V 2的含义是X 2的负载矩阵;P 1 T和P 2 T是Ps的两个子矩阵,Ps为S comb=[S 1,S 2]组合矩阵的负载矩阵;上标"T"表示转置。该方法能够实现近红外光谱和特征(或指纹)图谱之间的图谱转换,实现不同类型仪器之间的图谱转换。
Description
本发明要求于2021年9月22日提交中国专利局、申请号为202111106610.6、发明名称为“一种近红外光谱与特征图谱的图谱转换方法及其应用”的中国专利申请的优先权,其全部内容通过引用结合在本发明中。
本发明属于色谱分析技术领域,涉及一种近红外光谱与特征图谱的图谱转换方法及其应用。
公开该背景技术部分的信息仅仅旨在增加对本发明的总体背景的理解,而不必然被视为承认或以任何形式暗示该信息构成已经成为本领域一般技术人员所公知的现有技术。
近红外光谱分析方法(NIR)是当前发展迅速的一种无损、无污染、重现性好的分析技术,符合绿色分析化学技术(GAC)理念。随着化学计量学和计算机技术的发展,该技术已在农产品、石油化学、制药、环境、过程控制、临床及生物医学等领域广泛应用。该方法的一大特点是需要借助化学计量学将样品的光谱信息与对应的参考值信息(如含量、来源等)相关联建立模型,通过所建立的模型对未知的样品进行预测,从而实现分析的目的。近红外光谱快速、无损、绿色,适用于生产现场检测和实时在线分析,已成为过程分析技术(PAT)的重要组成部分,但近红外光谱的峰重叠严重,具体成分信息不明确。
指纹图谱系指某种中药材或中成药经适当处理后运用现代分析手段得到的能够表征或反映该中药特征的图谱或图像。特征图谱(或指纹)图谱技术在中药质量标准中的应用完善了中药复杂体系质量标准的研究思路、方法和模式,也是今后中药质量整体可控性的发展趋势。高效液相色谱(HPLC)分离效能高,特征强,可靠性好,适合作参考方法用于验证。但其预处理复杂、很难用于样品的直接分析,具有破坏样品原始形态、耗时长、污染环境等。因此色谱分析技术难以适用于中药生产过程及其中间物料的及时、快速分析,难以做大批次检验,通常用于中间物料和最终产品生产后的小样本检测及验证。
在近红外光谱分析领域,有经典的模型转移方法,如单变量斜率截距校正(SBC)法、光谱空间转换(SST)法、直接校正(DS)法和分段校正(PDS)法等等。其中SST法、DS法及PDS法都是基于光谱之间的转换实现模型转移。
中药的大部分有效成分均与含氢基团有关,而近红外光谱分析技术反映的正是含氢基团的全部信息,且中药物质成分可以采用指纹图谱进行测定,也可以用近红外光谱技术进行检测。 如果将快速的近红外(NIR)光谱测定法与分辨能力强且可靠的基于色谱技术的特征(或指纹)图谱方法相关联,实现由近红外光谱向特征(或指纹)图谱转化,既可以具有NIR光谱的快速、无损、绿色的特点,同时转化后的特征(或指纹)图谱也具有分辨能力强,可靠性好的特点,可以充分发挥两种分析技术的优势。但是,发明人发现,目前,关于NIR光谱与对应特征(或指纹)图谱的转化方法研究甚少,并没有行之有效的转化方法。
发明内容
为了解决现有技术的不足,本发明提供了一种近红外光谱与特征图谱的图谱转换方法及其应用,该方法能够实现近红外光谱和特征(或指纹)图谱之间的图谱转换,且两者的图谱变量数是否一致没有限制,能够将光谱和特征(或指纹)图谱建立联系,实现不同类型仪器之间的图谱转换。
具体地,本发明是通过如下技术方案实现的:
在本发明的第一方面,一种近红外光谱与特征图谱的图谱转换方法,包括以下步骤:首先,将光谱原始矩阵X
1和特征图谱原始矩阵X
2进行异常值剔除及预处理,然后进行奇异值分解,在保留相同的主成分数下,得到X
1的得分矩阵S
1和X
2的S
2;将S
1和S
2两个矩阵进行关联;通过公式X
2trans=[X
1V
1(P
1
T)
+(P
2
T)]V
2
T,将转换后的图谱校正为适合从仪器的特征图谱;其中,X
2trans表示转移后的特征图谱矩阵;V
1的含义是X
1的负载矩阵;V
2的含义是X
2的负载矩阵;P
1
T和P
2
T是Ps的两个子矩阵,Ps为S
comb=[S
1,S
2]组合矩阵的负载矩阵。上标“T”表示转置。
转换方向为由样本的近红外光谱向对应色谱特征(或指纹)图谱进行转换,根据后面所述算法,反向即由色谱特征(或指纹)图谱向NIR光谱转换理论上也是可行的。
在本发明的第二方面,任一所述的一种近红外光谱与特征图谱的图谱转换方法在中药成分检测中的应用。
本发明一个或多个实施例具有的有益效果:
(1)、所述方法能够实现近红外光谱和特征(或指纹)图谱之间的图谱转换(理论上可以相互转换),且两者的图谱变量数是否一致没有限制,能够将光谱和特征(或指纹)图谱建立联系,实现不同类型仪器之间的图谱转换,因此所述方法更加灵活,适用范围更加广泛。所述方法将中药色谱特征(或指纹)图谱的成分特征表征能力强的特点与近红外光谱的快速、无损等特点相结合,充分发挥两种分析技术的优势。
(2)、所述方法能较好的保留转换图谱原始数据的特征信息。
图1:实施例1中所有样本的原始近红外光谱;
图2:实施例1中所有样本的原始特征(或指纹)光谱;
图3:实施例1中近红外光谱去除异常样本的主成分投影图;
图4:实施例1中特征(或指纹)图谱去除异常样本的主成分投影图;
图5:实施例1从仪器预测集经本发明图谱转换前后的比较图谱;
图6:实施例1从仪器特征(或指纹)图谱选择的特征峰;
图7:实施例2中所有样本的原始近红外光谱;;
图8:实施例2中所有样本的原始特征(或指纹)光谱;
图9:实施例2中近红外光谱去除异常样本的主成分投影图;
图10:实施例2中特征(或指纹)图谱去除异常样本的主成分投影图;
图11:实施例2从仪器预测集经本发明图谱转换前后的比较图谱;
图12:实施例2从仪器特征(或指纹)图谱选择的特征峰。
下面结合具体实施例,进一步阐述本发明。应理解,这些实施例仅用于说明本发明而不用于限制本发明的范围。下列实施例中未注明具体条件的实验方法,通常按照常规条件或按照制造厂商所建议的条件。
需要注意的是,这里所使用的术语仅是为了描述具体实施方式,而非意图限制根据本发明的示例性实施方式。如在这里所使用的,除非上下文另外明确指出,否则单数形式也意图包括复数形式,此外,还应当理解的是,当在本说明书中使用术语“包含”和/或“包括”时,其指明存在特征、步骤、操作、器件、组件和/或它们的组合。
目前,关于NIR光谱与对应特征(或指纹)图谱的转化方法研究甚少,并没有行之有效的转化方法,为此,本发明提供了一种近红外光谱与特征图谱的图谱转换方法及其应用。
在本发明的一种或多种实施例中,一种近红外光谱与特征图谱的图谱转换方法,包括以下步骤:首先,将光谱原始矩阵X
1和特征图谱原始矩阵X
2进行异常值剔除及预处理,然后进行奇异值分解,在保留相同的主成分数下,得到X
1的得分矩阵S
1和X
2的S
2;将S
1和S
2两个矩阵进行关联;通过公式X
2trans=[X
1V
1(P
1
T)
+(P
2
T)]V
2
T,将转换后的图谱校正为适合从仪器的特征图谱;其中,X
2trans表示转移后的特征图谱矩阵;V
1的含义是X
1的负载矩阵;V
2的含义是X
2的负载矩阵;P
1
T和P
2
T是Ps的两个子矩阵,Ps为S
comb=[S
1,S
2]组合矩阵的负载矩阵;上标“T”表示转置。
转换方向为由样本的近红外光谱向对应色谱特征图谱进行转换,也可以反向,即由色谱特征图谱向近红外光谱转换,具体包括以下步骤:
(1)样本的光谱采集和特征图谱的测定:对样本分别进行主仪器的近红外光谱采集和从仪器的特征图谱测定,得光谱原始矩阵X
1和特征图谱原始矩阵X
2;
(2)异常值剔除:对光谱原始矩阵X
1和特征图谱原始矩阵X
2均进行异常值检测,将两者全部异常值剔除;
(3)光谱的预处理:对剔除异常值后的光谱原始矩阵X
1进行预处理,进行奇异值分解后,得到得分矩阵S
1;
(4)特征图谱的预处理:对剔除异常值后的从仪器采集的特征图谱原始矩阵X
2进行预处
理,得到特征图谱原始矩阵X
2;然后进行奇异值分解得到得分矩阵S
2;通过公式S
2=S
1–
S
1(P
1
T)
+(P
1
T-P
2
T)将S
1和S
2两个矩阵进行关联;
(5)样品集的划分:将预处理后的若干个标定样本划分为校正样本和预测样本,将主仪器的光谱原始矩阵X
1划分为校正集样本X
1mod和预测集样本X
1test;将从仪器的特征图谱原始矩阵X
2划分为校正集样本X
2trans_mod和预测集样本X
2trans_test;主仪器和从仪器的校正样本与预测样本要一一对应;
(6)图谱转换:将主仪器的光谱转换为从仪器的特征图谱,将转换后的图谱校正为适合从仪器的特征图谱。
(601)在图谱转换过程中,将两个矩阵进行组合,由于X
1和X
2不是相同类型仪器测定的图谱信号,X
1和X
2数据点数是不相同的,因此构建矩阵
S
1=X
1V
1 (1)
S
2=X
2V
2 (2)理论上说,只要能够通过式(1)和(2)的矩阵运算得到S
1和S
2即可。一种比较方便的运算方法是V
1和V
2分别为X
1和X
2主成分的负载矩阵,S
1和S
2则为对应的得分矩阵。因X
1和X
2测定的是相同的样本,因此X
1和X
2的主成分数是一致的。将S
1和S
2按照下式组合
S
comb=[S
1,S
2] (3)
将S
comb矩阵进行主成分分解,得到
S
comb=T
sP
s
T+E=T
s[P
1
T,P
2
T]+E (4)
其中,T
s与P
s分别为组合后S
comb矩阵的得分和负载矩阵;E表示相应的误差部分;上标“T”表示转置;下标“s”和“n”分别表示光谱信息和噪音的响应因子。主从仪器的样品数相同为S
comb矩阵的行数,P
1和P
2是P
s的两个子矩阵(P
s
T=[P
1
T,P
2
T])。
(602)S
1和S
2分别为X
1和X
2的得分矩阵,两者的差可以表示为S
1-S
2=T
s(P
1
T-P
2
T),因此,S
2可以通过下式进行计算:
S
2=S
1-T
s(P
1
T-P
2
T)=S
1–S
1(P
1
T)
+(P
1
T-P
2
T) (5)
其中,上标“+”表示矩阵的伪逆。
(603)将式(1)和(2)代入式(5)可以得到:
S
2=X
1V
1-X
1V
1(P
1
T)
+(P
1
T-P
2
T)=X
1V
1(P
1
T)
+P
2
T (6)
由式(2)可知,
X
2=S
2V
2
T (7)
因NIR图谱向特征(或指纹)图谱进行转换,所以有下式关系:
X
2=[X
1V
1(P
1
T)
+(P
2
T)]V
2
T (8)
如果若将特征(或指纹)图谱向NIR图谱转换,那么将X
1作为特征(或指纹)图谱,X
2作为NIR图谱即可。
(604)通过式(8)将主仪器校正集近红外光谱与从仪器特征(或指纹)图谱建立图谱转换关系模型:
X
2trans_mod=[X
1modV
1mod(P
1mod
T)
+(P
2mod
T)]V
2mod
T (9)
其中,V
1mod,P
1mod,P
2mod及V
2mod都是通过校正集样本得到的矩阵,与V
1,P
1,P
2及V
2的含义一一对应,即V
1mod和V
2mod分别为X
1mod(NIR光谱校正集样本)和X
2mod(特征(或指纹)图谱校正集样本)主成分的负载矩阵,P
1mod和P
2mod是P
smod的两个子矩阵,P
smod为校正集样本组合后S
comb_mod=[S
1mol,S
2mol]矩阵的负载矩阵,S
1mod与S
2mod分别是X
1mod与X
2mod的得分矩阵。对于未知待测样本的NIR光谱转换为对应的特征(或指纹)图谱按照下式进行:
X
2trans_test=[X
1testV
1mod(P
1mod
T)
+(P
2mod
T)]V
2mod
T (10)
优选的,步骤(2)中,采用Hotelling T
2法对异常值进行检测。
可选的,步骤(3)中,所述对光谱的预处理方式,包括:平滑处理、一阶导数计算、二阶导数计算、标准化处理、基线漂移处理、标准正态变量处理、多元散射校正处理等中的任意一种或多种的组合,也可以不使用预处理方法对光谱进行校正。
可选的,步骤(4)中,所述对特征(或指纹)图谱的预处理方法,包括相关优化翘曲法和自适应迭代加权最小二乘法等中的任意一种。
优选的,步骤(5)中,所述校正集的数量大于或等于预测集的数量,校正集和预测集样本数量的设置比例为2:1及以上。
可选的,步骤(5)中,所述若干个标定样本划分为校正样本和预测样本,划分方式包括:KS方法、Rank-KS方法、SPXY方法、Rank-SPXY方法及含量梯度法中的任意一种。
优选的,步骤(6)中,所述方法近红外光谱和特征(或指纹)图谱的样本数是必须相同 的,测定的样本要一一对应,但是光谱与色谱特征(或指纹)图谱的变量数可以是相等的,也可以不相等,适用范围和内容更广泛。
采用本发明所述方法能够将NIR光谱与特征(或指纹)图谱建立矩阵转换关系(如式(8),(9)和(10)),实现近红外光谱和特征(或指纹)图谱之间的图谱转换,且两者的图谱变量数是否一致没有限制,能够将光谱和特征(或指纹)图谱建立联系,实现不同类型仪器之间的图谱转换,因此本发明所述方法更加灵活,适用范围更加广泛。本发明所述方法将中药色谱特征(或指纹)图谱的成分特征表征能力强的特点与近红外光谱的快速、无损等特点相结合,充分发挥两种分析技术的优势。
在本发明的一种或多种实施例中,任一所述的一种近红外光谱与特征图谱的图谱转换方法在中药成分检测中的应用。
下面结合具体的实施例,对本发明做进一步的详细说明,应该指出,所述具体实施例是对本发明的解释而不是限定。
除非另有指明,本文使用的所有技术和科学术语具有与本申请所属技术领域的普通技术人员通常理解的相同含义。
具体实施例1:
以市售的参枝苓口服液为实例,测定的样本数共有17个。X
1为样本的近红外光谱矩阵,由作为主仪器的傅里叶变换近红外光谱仪(Antaris Ⅱ,赛默飞世尔,美国)测得,样本的原始光谱见图1。X
2为样本的特征(或指纹)图谱矩阵,由作为从仪器的高效液相色谱仪(Agilent 1260,安捷伦科技有限公司,美国)测得,样本的原始特征(或指纹)图谱见图2。
先进行异常样本的剔除,通过Hotelling T
2方法,没有检测到异常样本,去除异常值的主成分分析图见图3和图4。
对于主仪器,本实施事例中的样本光谱未使用预处理方法进行光谱处理。对于从仪器,样本特征(或指纹)图谱经相关优化翘曲(COW)方法进行校正,目的是使特征(或指纹)图谱的特征峰对齐。基于主仪器样本的光谱数据,采用常用的KS方法将样本集划分为校正集和预测集;校正集样本数为12,预测集样本是5。主仪器和从仪器的校正集样本和预测集样本的序号应保持一致。
图谱转换如下:首先需将光谱X
1和特征(或指纹)指纹图谱X
2两者均进行奇异值分解,在保留相同的主成分数下,得到S
1和S
2矩阵,分别为光谱的得分矩阵和特征(或指纹)图谱的得分矩阵;通过公式S
2=S
1–S
1(P
1
T)
+(P
1
T-P
2
T)将S
1和S
2两个矩阵进行关联。通过公式X
2trans=[X
1V
1(P
1
T)
+(P
2
T)]V
2
T,可以将转换后的图谱校正为适合从仪器的特征(或指纹)图谱, 转换前后的从仪器校正集图谱见图5,并对转换后的从仪器预测集样本进行预测,计算图谱转换前后预测集样本之间的相对分析误差(RPD)、预测均方根误差(RMSEP)和相似度(Similarity),以此评价图谱转换效果。
表1图谱转换前后特征(或指纹)图谱预测集结果
预测集样本序号 | 相似度 | RPD | RMSEP |
1 | 0.9922 | 4.87 | 6.0135 |
3 | 0.9801 | 4.39 | 6.1326 |
4 | 0.9963 | 5.16 | 4.7618 |
9 | 0.9978 | 6.90 | 3.3726 |
15 | 0.9976 | 10.98 | 2.3048 |
平均值 | 0.9928 | 6.46 | 4.5171 |
由表1可见,经过图谱转换后,预测集中所有样本经图谱转移后,其前后图谱的相似度均大于0.98,且其RPD值均大于4,表示转换后的模型效果是很好的。我们在图谱转换前后的特征(或指纹)图谱中选择同一个特征峰,所选特征峰见图6。计算该特征峰在图谱转换前后的峰面积以及两者之间的比值,在图谱转换前后峰面积的比值越接近1时,说明特征峰在转换前后的峰面积更为接近,图谱转换的效果更好。
表2图谱转换前后特征峰的峰面积结果
由表2可知,在五组预测集样品中,特征峰在转换前后其峰面积的比值在0.75至1.25之间,可以认为转换前后的结果是能够接受的,也进一步地说明参枝苓口服液通过本发明所述方法可以有很好的图谱转换效果。
具体实施例2:
以市售的心可舒片溶液为实例,测定的样本数共有25个。X
1为样本的近红外光谱矩阵,由作为主仪器的傅里叶变换近红外光谱仪(Antaris Ⅱ,赛默飞世尔,美国)测得,样本的原始光谱见图7。X
2为样本的特征(或指纹)图谱矩阵,由作为从仪器的高效液相色谱仪(Agilent 1260,安捷伦科技有限公司,美国)测得,样本的原始特征(或指纹)图谱见图8。
先进行异常样本的剔除,通过Hotelling T
2方法,检测到1个异常样本,剔除之后剩下24个样本,去除异常值的主成分分析图见图9和图10。
对于主仪器,本实施事例中的样本光谱采用标准正态变量变换(SNV)方法进行光谱预处理。对于从仪器,样本特征(或指纹)图谱经COW方法进行校正,目的是使特征(或指纹)图谱的特征峰对齐。基于主仪器样本的光谱数据,采用常用的Kennard-Stone(KS)方法将样本集划分为校正集和预测集;校正集样本数为14,预测集样本是10。主仪器和从仪器的校正集样本和预测集样本的序号应保持一致。
图谱转换如下:首先需将光谱X
1和特征(或指纹)指纹图谱X
2两者均进行奇异值分解,在保留相同的主成分数下,得到S
1和S
2矩阵,分别为光谱的得分矩阵和特征(或指纹)图谱的得分矩阵;通过公式S
2=S
1–S
1(P
1
T)
+(P
1
T-P
2
T)将S
1和S
2两个矩阵进行关联。通过公式X
2trans=[X
1V
1(P
1
T)
+(P
2
T)]V
2
T,可以将转换后的图谱校正为适合从仪器的特征(或指纹)图谱,转换前后的从仪器校正集图谱见图11,并对转换后的从仪器预测集样本进行预测,计算图谱转移前后预测集样本之间的相对分析误差(RPD)、预测均方根误差(RMSEP)和相似度(Similarity),以此评价图谱转换效果。
表3图谱转换前后特征(或指纹)图谱预测集结果
预测集样本序号 | 相似度 | RPD | RMSEP |
2 | 0.9927 | 7.84 | 0.7556 |
3 | 0.9889 | 6.28 | 0.8587 |
5 | 0.9928 | 3.86 | 1.7477 |
6 | 0.9930 | 7.81 | 0.7157 |
7 | 0.9816 | 3.31 | 1.4798 |
8 | 0.9581 | 3.06 | 1.9363 |
10 | 0.9496 | 2.96 | 2.0825 |
11 | 0.9432 | 2.59 | 2.3600 |
13 | 0.9957 | 8.99 | 0.6080 |
22 | 0.9789 | 4.48 | 1.1795 |
平均值 | 0.9774 | 5.12 | 1.3724 |
由表3可见,经过图谱转换后,所有预测集样本在图谱转换前后的相似度均大于0.94,说明心可舒片溶液有很好的图谱转移效果,并且所有样本的RPD值大于2.5,可以认为转换后的模型是可以接受的。我们在图谱转换前后的特征(或指纹)图谱中选择同一个特征峰,所选特征峰见图12。计算该特征峰在图谱转换前后的峰面积以及两者之间的比值,进一步评价图谱转换效果。
表4图谱转换前后特征峰的峰面积结果
由表4可知,在心可舒片溶液的10组预测集样品中,特征峰在转换前后其峰面积的比值基本上在0.75至1.25之间,且平均值为1.02接近于1,可以认为转换前后的结果是能够接受的,也进一步地说明心可舒片溶液通过本发明所述方法可以有很好的图谱转换效果。
由以上两个实例可以看出,但即使主仪器和从仪器两种图谱类型不同,主仪器近红外光谱经本发明所述图谱转换方法转换后得到的图谱与从仪器的特征(或指纹)图谱非常相似,说明本发明方法有效且效果较佳。所述图谱转换方法不局限于相同的图谱变量数,适用范围和内容更广泛。通过该方法,将中药色谱特征(或指纹)图谱(成分特征表征能力强)与近红外光谱(快速、无损等)相结合,可以充分发挥两种分析技术的优势。
以上所述仅为本申请的优选实施例而已,并不用于限制本申请,对于本领域的技术人员来说,本申请可以有各种更改和变化。凡在本申请的精神原则之内,所做的任何修改、等同替换、改进等,均应包含在本申请的保护范围之内。
Claims (10)
- 一种近红外光谱与特征图谱的图谱转换方法,其特征是,包括以下步骤:首先,将光谱原始矩阵X 1和特征图谱原始矩阵X 2进行异常值剔除及预处理,然后进行奇异值分解,在保留相同的主成分数下,得到X 1的得分矩阵S 1和X 2的S 2;将S 1和S 2两个矩阵进行关联;通过公式X 2trans=[X 1V 1(P 1 T) +(P 2 T)]V 2 T,将转换后的图谱校正为适合从仪器的特征图谱;其中,X 2trans表示转移后的特征图谱矩阵;V 1的含义是X 1的负载矩阵;V 2的含义是X 2的负载矩阵;P 1 T和P 2 T是Ps的两个子矩阵,Ps为S comb=[S 1,S 2]组合矩阵的负载矩阵;上标“T”表示转置。
- 如权利要求1所述的一种近红外光谱与特征图谱的图谱转换方法,其特征是,所述转换方法具体包括:(1)样本的光谱采集和特征图谱的测定:对样本分别进行主仪器的近红外光谱采集和从仪器的特征图谱测定,得光谱原始矩阵X 1和特征图谱原始矩阵X 2;(2)异常值剔除:对光谱原始矩阵X 1和特征图谱原始矩阵X 2均进行异常值检测,将两者全部异常值剔除;(3)光谱的预处理:对剔除异常值后的光谱原始矩阵X 1进行预处理,进行奇异值分解后,得到得分矩阵S 1;(4)特征图谱的预处理:对剔除异常值后的从仪器采集的特征图谱原始矩阵X 2进行预处理,得到特征图谱原始矩阵X 2;然后进行奇异值分解得到得分矩阵S 2;通过公式S 2=S 1–S 1(P 1 T) +(P 1 T-P 2 T)将S 1和S 2两个矩阵进行关联;(5)样品集的划分:将预处理后的若干个标定样本划分为校正样本和预测样本,将主仪器的光谱原始矩阵X 1划分为校正集样本X 1mod和预测集样本X 1test;将从仪器的特征图谱原始矩阵X 2划分为校正集样本X 2trans_mod和预测集样本X 2trans_test;主仪器和从仪器的校正样本与预测样本要一一对应;(6)图谱转换:将主仪器的光谱转换为从仪器的特征图谱,将转换后的图谱校正为适合从仪器的特征图谱。
- 如权利要求2所述的一种近红外光谱与特征图谱的图谱转换方法,其特征是,步骤(2)中,采用Hotelling T 2法对异常值进行检测。
- 如权利要求2所述的一种近红外光谱与特征图谱的图谱转换方法,其特征是,步骤(3)中,所述对光谱的预处理方式,包括:平滑处理、一阶导数计算、二阶导数计算、标准化处理、基线漂移处理、标准正态变量处理、多元散射校正处理中的任意一种或多种的组合。
- 如权利要求2所述的一种近红外光谱与特征图谱的图谱转换方法,其特征是,步骤(4)中,所述对特征图谱的预处理方法,包括相关优化翘曲法和自适应迭代加权最小二乘法中的任 意一种。
- 如权利要求2所述的一种近红外光谱与特征图谱的图谱转换方法,其特征是,步骤(5)中,所述校正集的数量大于或等于预测集的数量,校正集和预测集样本数量的设置比例为(5-2):1,优选的,为2:1。
- 如权利要求2所述的一种近红外光谱与特征图谱的图谱转换方法,其特征是,步骤(5)中,所述若干个标定样本划分为校正样本和预测样本,划分方式包括:KS方法、Rank-KS方法、SPXY方法、Rank-SPXY方法及含量梯度法中的任意一种。
- 如权利要求2所述的一种近红外光谱与特征图谱的图谱转换方法,其特征是,步骤(6)中,所述近红外光谱和特征图谱的样本数相同,测定的样本一一对应;或,步骤(6)中,SST法按照矩阵的行计算。
- 如权利要求2所述的一种近红外光谱与特征图谱的图谱转换方法,其特征是,步骤(6)中,所述图谱转换的方法包括:将主仪器校正集X 1mod和预测集X 1test的光谱进行计算分别得到适合从仪器校正集光谱X 2trans_mod和预测集光谱X 2trans_test:X 2trans_mod=[X 1modV 1(P 1 T) +(P 2 T)]V 2 TX 2trans_test=[X 1testV 1(P 1 T) +(P 2 T)]V 2 T其中,V 1的含义是X 1mod的负载矩阵;V 2的含义是X 2mod的负载矩阵;P 1 T和P 2 T是Ps的两个子矩阵,Ps为基于校正集样本的S comb=[S 1,S 2]组合矩阵的负载矩阵;上标“T”表示转置。
- 权利要求1-9任一所述的一种近红外光谱与特征图谱的图谱转换方法在中药成分检测中的应用。
Applications Claiming Priority (2)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN202111106610.6 | 2021-09-22 | ||
CN202111106610.6A CN113762208B (zh) | 2021-09-22 | 2021-09-22 | 一种近红外光谱与特征图谱的图谱转换方法及其应用 |
Publications (1)
Publication Number | Publication Date |
---|---|
WO2023046216A1 true WO2023046216A1 (zh) | 2023-03-30 |
Family
ID=78796743
Family Applications (1)
Application Number | Title | Priority Date | Filing Date |
---|---|---|---|
PCT/CN2022/132598 WO2023046216A1 (zh) | 2021-09-22 | 2022-11-17 | 一种近红外光谱与特征图谱的图谱转换方法及其应用 |
Country Status (2)
Country | Link |
---|---|
CN (1) | CN113762208B (zh) |
WO (1) | WO2023046216A1 (zh) |
Families Citing this family (4)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN108509997A (zh) * | 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) | 一种基于近红外光谱技术对中药皂角刺的真伪进行化学模式识别的方法 |
CN113762208B (zh) * | 2021-09-22 | 2023-07-28 | 山东大学 | 一种近红外光谱与特征图谱的图谱转换方法及其应用 |
CN114993982A (zh) * | 2022-06-02 | 2022-09-02 | 震坤行工业超市(上海)有限公司 | 油液性能参数的计算方法及在线监测润滑油的装置 |
CN115128006A (zh) * | 2022-07-22 | 2022-09-30 | 山东大学 | 一种全自动中药口服溶液质量评价系统及方法 |
Citations (5)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
US5459677A (en) * | 1990-10-09 | 1995-10-17 | Board Of Regents Of The University Of Washington | Calibration transfer for analytical instruments |
US5710713A (en) * | 1995-03-20 | 1998-01-20 | The Dow Chemical Company | Method of creating standardized spectral libraries for enhanced library searching |
CN109444066A (zh) * | 2018-10-29 | 2019-03-08 | 山东大学 | 基于光谱数据的模型转移方法 |
CN111563436A (zh) * | 2020-04-28 | 2020-08-21 | 东北大学秦皇岛分校 | 一种基于ct-cdd的红外光谱测量仪器标定迁移方法 |
CN113762208A (zh) * | 2021-09-22 | 2021-12-07 | 山东大学 | 一种近红外光谱与特征图谱的图谱转换方法及其应用 |
Family Cites Families (4)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
US20050250212A1 (en) * | 2004-05-07 | 2005-11-10 | Hormoz Azizian | FT-NIR fatty acid determination method |
CN101231274B (zh) * | 2008-01-28 | 2011-06-01 | 河南中医学院 | 近红外光谱快速测定山药中尿囊素含量的方法 |
CN105928901B (zh) * | 2016-07-11 | 2019-06-07 | 上海创和亿电子科技发展有限公司 | 一种定性定量相结合的近红外定量模型构建方法 |
CN110687072B (zh) * | 2019-10-17 | 2020-12-01 | 山东大学 | 一种基于光谱相似度的校正集和验证集的选择及建模方法 |
-
2021
- 2021-09-22 CN CN202111106610.6A patent/CN113762208B/zh active Active
-
2022
- 2022-11-17 WO PCT/CN2022/132598 patent/WO2023046216A1/zh unknown
Patent Citations (5)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
US5459677A (en) * | 1990-10-09 | 1995-10-17 | Board Of Regents Of The University Of Washington | Calibration transfer for analytical instruments |
US5710713A (en) * | 1995-03-20 | 1998-01-20 | The Dow Chemical Company | Method of creating standardized spectral libraries for enhanced library searching |
CN109444066A (zh) * | 2018-10-29 | 2019-03-08 | 山东大学 | 基于光谱数据的模型转移方法 |
CN111563436A (zh) * | 2020-04-28 | 2020-08-21 | 东北大学秦皇岛分校 | 一种基于ct-cdd的红外光谱测量仪器标定迁移方法 |
CN113762208A (zh) * | 2021-09-22 | 2021-12-07 | 山东大学 | 一种近红外光谱与特征图谱的图谱转换方法及其应用 |
Non-Patent Citations (1)
Title |
---|
HU, LIPING: "Study on Model Building and Model Transfer in the Application of near Infrared Spectroscopy to Online Quality Control of Chinese Herbal Liquid Tonic", CHINA MASTER’S THESES FULL-TEXT DATABASE, ENGINEERING SCIENCE AND TECHNOLOGY I, no. 3, 19 April 2020 (2020-04-19), China, pages 1 - 53, XP009544876, DOI: 10.27151/d.cnki.ghnlu.2020.004695 * |
Also Published As
Publication number | Publication date |
---|---|
CN113762208A (zh) | 2021-12-07 |
CN113762208B (zh) | 2023-07-28 |
Similar Documents
Publication | Publication Date | Title |
---|---|---|
WO2023046216A1 (zh) | 一种近红外光谱与特征图谱的图谱转换方法及其应用 | |
CN110687072B (zh) | 一种基于光谱相似度的校正集和验证集的选择及建模方法 | |
CN109444066B (zh) | 基于光谱数据的模型转移方法 | |
CN109187614B (zh) | 基于核磁共振和质谱的代谢组学数据融合方法及其应用 | |
CN111157698A (zh) | 一种利用发射率数据获取黑土土壤全钾含量的反演方法 | |
Fan et al. | Direct calibration transfer to principal components via canonical correlation analysis | |
CN113588847B (zh) | 一种生物代谢组学数据处理方法、分析方法及装置和应用 | |
CN108169165B (zh) | 基于太赫兹光谱和图像信息融合的麦芽糖混合物定量分析方法 | |
CN110503156B (zh) | 一种基于最小相关系数的多变量校正特征波长选择方法 | |
CN102313712B (zh) | 一种纤维类物料不同分光方式近红外光谱差异的校正方法 | |
CN114611582B (zh) | 一种基于近红外光谱技术分析物质浓度的方法及系统 | |
CN111896497B (zh) | 一种基于预测值的光谱数据修正方法 | |
WO2023207453A1 (zh) | 一种基于光谱聚类的中药成分分析方法及系统 | |
CN117269106A (zh) | 快速预测辣椒蛋白质含量的高光谱模型 | |
CN115144362A (zh) | 基于标准校准板的光谱模型自适应方法 | |
CN111220565B (zh) | 一种基于cpls的红外光谱测量仪器标定迁移方法 | |
CN109272561B (zh) | 基于空谱联合多假设预测的高光谱图像压缩感知重构方法 | |
CN108414475B (zh) | 基于光学层析同时迭代重建的libs分析方法 | |
CN117647498A (zh) | 一种牛奶的中红外光谱不同仪器标准化参数的获取方法 | |
CN109406419B (zh) | 基于高光谱成像技术预测枸杞子中对羟基苯甲酸含量的方法 | |
CN109406421B (zh) | 基于高光谱成像技术预测枸杞子中阿魏酸含量的方法 | |
CN109406420B (zh) | 基于高光谱成像技术预测枸杞子中东莨菪内酯含量的方法 | |
Li et al. | Robust Multi-task Learning for Calibration Transfer in DP Detection by NIRS of Insulating Paper | |
CN115795225B (zh) | 一种近红外光谱校正集的筛选方法及装置 | |
CN117571685A (zh) | 一种生物制药领域拉曼光谱跨设备模型转移方法 |
Legal Events
Date | Code | Title | Description |
---|---|---|---|
121 | Ep: the epo has been informed by wipo that ep was designated in this application |
Ref document number: 22872259 Country of ref document: EP Kind code of ref document: A1 |
|
NENP | Non-entry into the national phase |
Ref country code: DE |