论文部分内容阅读
Abstract This study was conducted to establish a method for rapid determination of crude starch content in complete quinoa （Chenopodium quinoa Willd） seeds. The near infrared spectra of 100 quinoa samples were collected， and a mathematic model was built using the near infrared spectra within the wavelength range of 1 0 000-4 000 cm-1 by first derivative +vector normalization spectral pre??processing. The results showed that the quantification model of starch content had better calibration and prediction effects， and showed a determination coefficient of cross validation （r2cv） of 0.914 7 and a determination coefficient of validation （r2val） of 0.903 1. The determination of starch content in complete quinoa seeds by near infrared spectroscopy is totally feasible.
Key words Chenopodium quinoa Willd.； Starch； Near infrared spectroscopy
Quinoa （Chenopodium quinoa Willd.） is native to Indian areas. It is a kind of annual herb with a planting history of 5 000-7 000 years， as well as a "false cereal" capable of growing well under severe environment （salt and alkali， drought， frost， and diseases and pests）[1-2]. Due to the full nutritive value and unique functional characteristics of quinoa， United Nations Food and Agriculture Organization affirms it as one and only perfect nutritious food， and United Nations General Assembly also announced in 2013 that the year is "the international quinoa year"[3-4]. Researches show that abundant protein， balanced amino acid composition and rich starch， fat， mineral and vitamin in quinoa could satisfy the demands for nutritive elements by human body[5]. Long??term consumption of quinoa can well treat heart disease， high blood pressure， high blood sugar and high blood lipid[6]. Shen et al.[7] found that compared with common cereals such as wheat， rice and millet， quinoa has a lower starch content， and is thus also suitable for diabetic and people on diet[7]. Among the carbohydrates in quinoa， the starch content is the highest， and accounts for 58.0%-64.2% of dry matter[8]. Quinoa starch has typical A type X diffraction structure， with strong stability during freezing and aging process， and its gelatinization point was 64 ??[9].
The active biological membrane has stronger antibacterial activity against 99% of Escherichia coli and 98% of Staphylococcus aureus， and is used for prolonging shelf life in food packaging[10]. Therefore， the research and development of quinoa starch attracts extensive attention. Shen et al.[7] determined the starch content in quinoa to be in the range of 52.28%-61.85% by dual??wavelength method. In addition， crude starch content in quinoa also could be determined using a polarimeter. Conventional methods for the determination of starch content suffer from the problems including tedious steps， low determination speed， high cost， long cycle and pulverization of grains， resulting in some unnecessary waste. Due to that quinoa germplasm resources and parent materials are much precious， and the seeds of some resources are relatively less， conventional methods do not satisfy the requirements for rapidity and completeness in the utilization of germplasm and parent materials， thereby greatly reducing their utilization efficiency. Therefore， it is urgent to develop a kind of rapid accurate detection method without the need for pre??treatment， which is beneficial to the improvement of efficiency of quinoa breeding for quality and the acceleration of breeding process. Near infrared spectroscopy （NIRS） is a spectroscopic analysis method developing the fast attracting the most attention since the 1990s， which has distinct advantages of high speed， high efficiency， simple sample preparation and no pollution[12]. The near infrared spectral region is 780-2 500 nm， which has strong response to the vibration of groups including C??H， N??H and O??H， and could reflect abundant material composition information and structural information. In recent years， NIRS means has made good achievements in the detection of agricultural crops[13-15]. In this study， the moisture and crude starch contents in 100 quinoa materials were determined by chemical methods， and then divided into calibration and validation sets， and a rapid detection and pre??processing NIRS model of starch was built， providing technique support for further rapid detection and utilization of quinoa resources.
Materials and Methods
Materials and instruments
Tested materials： The tested 100 quinoa varieties were provided by researchers Lu Ping and Tao Mei from the introduction department， Institute of Crop Sciences， Chinese Academy of Agricultural Sciences， and Institute of Crop Germplasm Resources， Shanxi Academy of Agricultural Sciences. The reagents required by the experiment included anhydrous ethanol （Tianjin Zhiyuan Chemical Reagent Co.， Ltd.）， hydrochloric acid （Beijing Chemical Works）， zinc sulfate （Tianjin Kaitong Chemical Reagent Co.， Ltd.） and potassium ferrocyanide （Tianjin Kaitong Chemical Reagent Co.， Ltd.）， all of which were analytically pure.
Instruments： Analytical balance （BSA124S Sartorius company）； cyclone mill Cyclotec1093 （Foss， Denmark）； electro??thermostatic blast oven （Ningbo Dongnan Instruments Co.， Ltd.）； MPA Fourier Transform Infrared Spectrometer （Bruke， Germany）； AP??300 polarimeter （Shanghai Shuangxu Electronics Co.， Ltd.）.
Experimental method
Determination method of crude starch content of quinoa With reference to GB 5006??1985 "Determination Methods of Crude Starch in Grains"， two parts of the same sample were weighed， each of which was 2.5 g， accurate to 0.001 g. The crude starch content was measured with a polarimeter.
Acquisition of near infrared spectra In order to acquire an optimal model and prediction effect， the collected quinoa samples were stored under room temperature for about one week to balance moisture， and meanwhile， impurities in the samples and the grains with obviously different shapes were removed （one part was subjected to shelling treatment， and the other part was subjected to shelling and pulverization and sieved with a 60 mesh sieve）. At first， the near infrared spectrometer was preheated for 30 min， and performance test and adjustment of baseline were performed， followed by determination of samples. The MPA Fourier Transform Infrared Spectrometer from Bruke company was operated to perform diffuse reflection spectrum scanning， and the working spectrum range was selected as 4 000-12 000 cm-1. Each sample was loaded and scanned for two times， and the values were averaged. The reflection spectrum information was automatically converted to absorbance values which were stored， and the experimental data were analyzed on OPUS modeling software. Construction of near infrared mathematical model Spectral data pre??processing， spectral range selection and regression statistical analysis were performed using OPUS/QUAN T 5.5 quantitative spectrometric analysis software provided by Bruker company and DPS software. According to the requirements to modeling data， the data of 100 quinoa samples were grouped， 80% of which were used for the construction of near infrared model， and 20% were used for checking the precision of the constructed model， serving as the validation set[16]. In order to find an optimal modeling method， different modeling methods were used for the building of quantitative models of main components in quinoa. Internal validation was performed with the calibration sample set at first， followed by external validation of the models with non??modeling samples selected randomly， to investigate the adaptability and precision of models， i.e.， the optimal model was determined according to the indexes including determination coefficient of calibration r2cal， root mean square error of estimation RMSEE， determination coefficients of cross validation r2cv， root mean square error of cross validation RMSECV， determination coefficient of validation r2val and root mean square error of prediction RMSEP.
Results and Analysis
Quinoa original spectrum and chemical values
The original spectra of the 100 quinoa samples are shown in Fig. 1. It could be seen that quinoa has multiple absorption peaks in the range of 10 000-4 000 cm-1， and their changing trends are accordant but not coincident.
The starch contents of the 100 quinoa samples are shown in Table 1， including 80 in the calibration set and 20 in the validation set. The crude starch contents were in the range of 49.97%-59.32%， and the average value of the three replicates was 54.63%. The data have a wider changing range， and are suitable for the building of near infrared analysis model， with better applicability.
Building of quinoa starch model In this study， the automatic optimization function in OPUS/QUAN T5.5 software was applied to screen the optimal spectral pre??processing method， spectral range and main factor number for modeling. Through cross validation， the parameters including the determination coefficient of cross validation r2 and root mean square error of cross validation RMSECV were compared under different combinations of spectral pre??processing method and spectral range， so as to determine the optimal calibration model. Fig. 2 shows that the calibration model of starch content built by first derivative +vector normalization spectral pre??processing had a better calibration effect， and showed a determination coefficient of cross validation （r2cv） of 91.47 and root mean square error of cross validation （RMSECV） of 0.481. External validation of quinoa starch model
The samples in the completely independent validation set with known chemical components， which did not participate in the building of the model， were used to evaluate the quality and actual prediction effect of the constructed model. The determination coefficient of validation （r2val） was 0.903 1， and the root mean square error of prediction （RMSEP） was 0.518. There were no significant differences between the prediction values of the model and the actual values. The starch contents obtained by the two methods were basically accordant， indicating that the results determined by the near infrared quality analyzer are accurate and reliable.
Discussion
The conventional method for the determination of starch content is polarimetry method， which is time and labor consuming with a long period， while near infrared analysis is a kind of indirect detection taking the chemical data tested in laboratory as a basis， which， as a new detection method， has the advantages of fast analysis， simultaneously determination of multiple components， low analysis cost and simple operation， and attracts more and more attention from researchers. In this study， 100 ordinary quinoa samples were selected， and a prediction model of starch contents in complete quinoa grains was constructed by NIRS analysis technique. The built model has higher determination coefficient （r2=91.47） and small error. In breeding for quinoa resources and quality， the model could rapidly complete the identification and analysis of resources and breeding materials， and achieve the effect of greatly shortening working period， reducing workload and promoting rapid identification and application of resources， which is of certain significance to the improvement of breeding efficiency.
The materials used in this study have better representativeness， but the changing range of starch content in the samples is still not wide enough. Furthermore， during the process of model construction， some abnormal values of starch contents of some resources were removed， which affects the accuracy of the rapid detection model to a certain degree. Therefore， it is necessary to continuously increase quinoa resource number to widen the changing ranges of related components in the samples， so as to cover the changing ranges of related components in quinoa production and breeding materials. Meanwhile， further optimization of the rapid detection model should be performed， so as to improve its accuracy and utilization efficiency. Agricultural Biotechnology 2018References
[1] JACOBSEN SE， MUJICA A， JENSEN CR. The resistance of quinoa （Chenopodium quinoa Willd.） to adverse abiotic factors[J]. Food Reviews International， 2003， 19（1/2）： 99-109.
[2] IQBAL MJ， REDDY OUK， EL??ZIK KM， et al. A genetic bottleneck in the ??envolution under domestication?? of upland cotton Gossypium hirsutum L. examined using DNA fingerprinting[J]. Theoretical and Applied Genetics， 2001， 103（4）： 547-554.
[3] COMAI S， BERTAZZO A， BAILONI L， et al. The content of proteic and nonproteic （free and protein??bound） tryptophan in quinoa and cereal flours[J]. Food Chemistry， 2007， 100（4） ： 1350-1355.
[4] OSHODI AA， OGUNGBENLE HN， OLADIMEJI MO. Chemical composition， nutritionally valuable minerals and functional properties of benniseed （Sesamun radiatum）， pearl millet （Pennisetum typhoides） and quinoa （Chenopodium quinoa ） flours[J]. International Journal of Food Sciences and Nutrition， 1999， 50（5）： 325-331.
[5] VEGA??G?BLVEZ A， MIRANDA M， VERGARA J， et al. Nutrition facts and functional potential of quinoa （Chenopodium quinoa Willd.） ， an ancient Andean grain： a review[J]. Journal of the Science of Food and Agriculture， 2010， 90（15）： 2541-2547.
[6] XIAO ZC， ZHANG GL. Development and utilization of Chenopodium quinoa Willd.[J]. Chinese Wild Plant Resources， 2014， 33（2） ： 62-66.
[7] SHEN RL， ZHANG WJ， DONG JL， et al. Determination of main nutritional component， mineral element and phytochemical in Chenopodium quinoa Willd.[J]. Journal of Zhengzhou University of Light Industry： Natural Science， 2015， 30（5） ： 17-21.
[8] REPO??CARRASCO R， ESPINOZA C， JACOBSEN SE. Nutritional value and use of the Andean crops quinoa （Chenopodium quinoa） and kaiwa （Chenopodium pallidicaule） [J]. Food Reviews International， 2003， 19（1/2）： 179-189.
[9] NIENKE L. Studies on the characterization， biosynthesis and isolation of starch and protein from quinoa （Chenopodium quinoa Willd.）[D]. Saskatchewan， Canada： University of Saskatchewan， 2005.
[10] PAGNO CH， COSTA TM， DE MENEZES EW， et al. Development of active biofilms of quinoa （Chenopodium quinoa W.） starch containing gold nanoparticles and evaluation of antimicrobial activity[J]. Food Chemistry， 2015， 173（173） ： 755-762.
[11] ZHOU HT， LIU H， YAO Y， et al. Evaluation of agronomic and quality characters of quinoa cultivated in Zhangjiakou [J]. Journal of Plant Genetic Resources， 2014， 15（1）： 222-227.
[12] YAN YL， ZHAO LL， LI JH， et al. Information technology of modern NIR spectral analysis[J]. Spectroscopy and Spectral Analysis， 2000， 20（6）： 777-780.
[13] DELWICHE SR， HRUSCHKA WR. Protein content of bulk wheat from near??infrared reflectance of individual kernels [J]. Cereal Chemistry， 2000， 77（1）： 86-88.
[14] YAN YL， ZHAO LL， HAN DH， et al. Near infrared spectroscopic analysis basis and application[M]. Beijing： China Light Industry Press， 2005： 3-10.
[15] YAN YL， CHEN B， ZHU DZ， et al. Principle， technique and application of near infrared spectroscopy[M]. Beijing： China Light Industry Press， 2013： 1-7.
[16] XU GT， YUAN HF， LU WZ. General situation and progress of near??infrared spectroscopy analysis instrument[J]. Modern Scientific Instruments， 1997（3） ： 9-11.
Key words Chenopodium quinoa Willd.； Starch； Near infrared spectroscopy
Quinoa （Chenopodium quinoa Willd.） is native to Indian areas. It is a kind of annual herb with a planting history of 5 000-7 000 years， as well as a "false cereal" capable of growing well under severe environment （salt and alkali， drought， frost， and diseases and pests）[1-2]. Due to the full nutritive value and unique functional characteristics of quinoa， United Nations Food and Agriculture Organization affirms it as one and only perfect nutritious food， and United Nations General Assembly also announced in 2013 that the year is "the international quinoa year"[3-4]. Researches show that abundant protein， balanced amino acid composition and rich starch， fat， mineral and vitamin in quinoa could satisfy the demands for nutritive elements by human body[5]. Long??term consumption of quinoa can well treat heart disease， high blood pressure， high blood sugar and high blood lipid[6]. Shen et al.[7] found that compared with common cereals such as wheat， rice and millet， quinoa has a lower starch content， and is thus also suitable for diabetic and people on diet[7]. Among the carbohydrates in quinoa， the starch content is the highest， and accounts for 58.0%-64.2% of dry matter[8]. Quinoa starch has typical A type X diffraction structure， with strong stability during freezing and aging process， and its gelatinization point was 64 ??[9].
The active biological membrane has stronger antibacterial activity against 99% of Escherichia coli and 98% of Staphylococcus aureus， and is used for prolonging shelf life in food packaging[10]. Therefore， the research and development of quinoa starch attracts extensive attention. Shen et al.[7] determined the starch content in quinoa to be in the range of 52.28%-61.85% by dual??wavelength method. In addition， crude starch content in quinoa also could be determined using a polarimeter. Conventional methods for the determination of starch content suffer from the problems including tedious steps， low determination speed， high cost， long cycle and pulverization of grains， resulting in some unnecessary waste. Due to that quinoa germplasm resources and parent materials are much precious， and the seeds of some resources are relatively less， conventional methods do not satisfy the requirements for rapidity and completeness in the utilization of germplasm and parent materials， thereby greatly reducing their utilization efficiency. Therefore， it is urgent to develop a kind of rapid accurate detection method without the need for pre??treatment， which is beneficial to the improvement of efficiency of quinoa breeding for quality and the acceleration of breeding process. Near infrared spectroscopy （NIRS） is a spectroscopic analysis method developing the fast attracting the most attention since the 1990s， which has distinct advantages of high speed， high efficiency， simple sample preparation and no pollution[12]. The near infrared spectral region is 780-2 500 nm， which has strong response to the vibration of groups including C??H， N??H and O??H， and could reflect abundant material composition information and structural information. In recent years， NIRS means has made good achievements in the detection of agricultural crops[13-15]. In this study， the moisture and crude starch contents in 100 quinoa materials were determined by chemical methods， and then divided into calibration and validation sets， and a rapid detection and pre??processing NIRS model of starch was built， providing technique support for further rapid detection and utilization of quinoa resources.
Materials and Methods
Materials and instruments
Tested materials： The tested 100 quinoa varieties were provided by researchers Lu Ping and Tao Mei from the introduction department， Institute of Crop Sciences， Chinese Academy of Agricultural Sciences， and Institute of Crop Germplasm Resources， Shanxi Academy of Agricultural Sciences. The reagents required by the experiment included anhydrous ethanol （Tianjin Zhiyuan Chemical Reagent Co.， Ltd.）， hydrochloric acid （Beijing Chemical Works）， zinc sulfate （Tianjin Kaitong Chemical Reagent Co.， Ltd.） and potassium ferrocyanide （Tianjin Kaitong Chemical Reagent Co.， Ltd.）， all of which were analytically pure.
Instruments： Analytical balance （BSA124S Sartorius company）； cyclone mill Cyclotec1093 （Foss， Denmark）； electro??thermostatic blast oven （Ningbo Dongnan Instruments Co.， Ltd.）； MPA Fourier Transform Infrared Spectrometer （Bruke， Germany）； AP??300 polarimeter （Shanghai Shuangxu Electronics Co.， Ltd.）.
Experimental method
Determination method of crude starch content of quinoa With reference to GB 5006??1985 "Determination Methods of Crude Starch in Grains"， two parts of the same sample were weighed， each of which was 2.5 g， accurate to 0.001 g. The crude starch content was measured with a polarimeter.
Acquisition of near infrared spectra In order to acquire an optimal model and prediction effect， the collected quinoa samples were stored under room temperature for about one week to balance moisture， and meanwhile， impurities in the samples and the grains with obviously different shapes were removed （one part was subjected to shelling treatment， and the other part was subjected to shelling and pulverization and sieved with a 60 mesh sieve）. At first， the near infrared spectrometer was preheated for 30 min， and performance test and adjustment of baseline were performed， followed by determination of samples. The MPA Fourier Transform Infrared Spectrometer from Bruke company was operated to perform diffuse reflection spectrum scanning， and the working spectrum range was selected as 4 000-12 000 cm-1. Each sample was loaded and scanned for two times， and the values were averaged. The reflection spectrum information was automatically converted to absorbance values which were stored， and the experimental data were analyzed on OPUS modeling software. Construction of near infrared mathematical model Spectral data pre??processing， spectral range selection and regression statistical analysis were performed using OPUS/QUAN T 5.5 quantitative spectrometric analysis software provided by Bruker company and DPS software. According to the requirements to modeling data， the data of 100 quinoa samples were grouped， 80% of which were used for the construction of near infrared model， and 20% were used for checking the precision of the constructed model， serving as the validation set[16]. In order to find an optimal modeling method， different modeling methods were used for the building of quantitative models of main components in quinoa. Internal validation was performed with the calibration sample set at first， followed by external validation of the models with non??modeling samples selected randomly， to investigate the adaptability and precision of models， i.e.， the optimal model was determined according to the indexes including determination coefficient of calibration r2cal， root mean square error of estimation RMSEE， determination coefficients of cross validation r2cv， root mean square error of cross validation RMSECV， determination coefficient of validation r2val and root mean square error of prediction RMSEP.
Results and Analysis
Quinoa original spectrum and chemical values
The original spectra of the 100 quinoa samples are shown in Fig. 1. It could be seen that quinoa has multiple absorption peaks in the range of 10 000-4 000 cm-1， and their changing trends are accordant but not coincident.
The starch contents of the 100 quinoa samples are shown in Table 1， including 80 in the calibration set and 20 in the validation set. The crude starch contents were in the range of 49.97%-59.32%， and the average value of the three replicates was 54.63%. The data have a wider changing range， and are suitable for the building of near infrared analysis model， with better applicability.
Building of quinoa starch model In this study， the automatic optimization function in OPUS/QUAN T5.5 software was applied to screen the optimal spectral pre??processing method， spectral range and main factor number for modeling. Through cross validation， the parameters including the determination coefficient of cross validation r2 and root mean square error of cross validation RMSECV were compared under different combinations of spectral pre??processing method and spectral range， so as to determine the optimal calibration model. Fig. 2 shows that the calibration model of starch content built by first derivative +vector normalization spectral pre??processing had a better calibration effect， and showed a determination coefficient of cross validation （r2cv） of 91.47 and root mean square error of cross validation （RMSECV） of 0.481. External validation of quinoa starch model
The samples in the completely independent validation set with known chemical components， which did not participate in the building of the model， were used to evaluate the quality and actual prediction effect of the constructed model. The determination coefficient of validation （r2val） was 0.903 1， and the root mean square error of prediction （RMSEP） was 0.518. There were no significant differences between the prediction values of the model and the actual values. The starch contents obtained by the two methods were basically accordant， indicating that the results determined by the near infrared quality analyzer are accurate and reliable.
Discussion
The conventional method for the determination of starch content is polarimetry method， which is time and labor consuming with a long period， while near infrared analysis is a kind of indirect detection taking the chemical data tested in laboratory as a basis， which， as a new detection method， has the advantages of fast analysis， simultaneously determination of multiple components， low analysis cost and simple operation， and attracts more and more attention from researchers. In this study， 100 ordinary quinoa samples were selected， and a prediction model of starch contents in complete quinoa grains was constructed by NIRS analysis technique. The built model has higher determination coefficient （r2=91.47） and small error. In breeding for quinoa resources and quality， the model could rapidly complete the identification and analysis of resources and breeding materials， and achieve the effect of greatly shortening working period， reducing workload and promoting rapid identification and application of resources， which is of certain significance to the improvement of breeding efficiency.
The materials used in this study have better representativeness， but the changing range of starch content in the samples is still not wide enough. Furthermore， during the process of model construction， some abnormal values of starch contents of some resources were removed， which affects the accuracy of the rapid detection model to a certain degree. Therefore， it is necessary to continuously increase quinoa resource number to widen the changing ranges of related components in the samples， so as to cover the changing ranges of related components in quinoa production and breeding materials. Meanwhile， further optimization of the rapid detection model should be performed， so as to improve its accuracy and utilization efficiency. Agricultural Biotechnology 2018References
[1] JACOBSEN SE， MUJICA A， JENSEN CR. The resistance of quinoa （Chenopodium quinoa Willd.） to adverse abiotic factors[J]. Food Reviews International， 2003， 19（1/2）： 99-109.
[2] IQBAL MJ， REDDY OUK， EL??ZIK KM， et al. A genetic bottleneck in the ??envolution under domestication?? of upland cotton Gossypium hirsutum L. examined using DNA fingerprinting[J]. Theoretical and Applied Genetics， 2001， 103（4）： 547-554.
[3] COMAI S， BERTAZZO A， BAILONI L， et al. The content of proteic and nonproteic （free and protein??bound） tryptophan in quinoa and cereal flours[J]. Food Chemistry， 2007， 100（4） ： 1350-1355.
[4] OSHODI AA， OGUNGBENLE HN， OLADIMEJI MO. Chemical composition， nutritionally valuable minerals and functional properties of benniseed （Sesamun radiatum）， pearl millet （Pennisetum typhoides） and quinoa （Chenopodium quinoa ） flours[J]. International Journal of Food Sciences and Nutrition， 1999， 50（5）： 325-331.
[5] VEGA??G?BLVEZ A， MIRANDA M， VERGARA J， et al. Nutrition facts and functional potential of quinoa （Chenopodium quinoa Willd.） ， an ancient Andean grain： a review[J]. Journal of the Science of Food and Agriculture， 2010， 90（15）： 2541-2547.
[6] XIAO ZC， ZHANG GL. Development and utilization of Chenopodium quinoa Willd.[J]. Chinese Wild Plant Resources， 2014， 33（2） ： 62-66.
[7] SHEN RL， ZHANG WJ， DONG JL， et al. Determination of main nutritional component， mineral element and phytochemical in Chenopodium quinoa Willd.[J]. Journal of Zhengzhou University of Light Industry： Natural Science， 2015， 30（5） ： 17-21.
[8] REPO??CARRASCO R， ESPINOZA C， JACOBSEN SE. Nutritional value and use of the Andean crops quinoa （Chenopodium quinoa） and kaiwa （Chenopodium pallidicaule） [J]. Food Reviews International， 2003， 19（1/2）： 179-189.
[9] NIENKE L. Studies on the characterization， biosynthesis and isolation of starch and protein from quinoa （Chenopodium quinoa Willd.）[D]. Saskatchewan， Canada： University of Saskatchewan， 2005.
[10] PAGNO CH， COSTA TM， DE MENEZES EW， et al. Development of active biofilms of quinoa （Chenopodium quinoa W.） starch containing gold nanoparticles and evaluation of antimicrobial activity[J]. Food Chemistry， 2015， 173（173） ： 755-762.
[11] ZHOU HT， LIU H， YAO Y， et al. Evaluation of agronomic and quality characters of quinoa cultivated in Zhangjiakou [J]. Journal of Plant Genetic Resources， 2014， 15（1）： 222-227.
[12] YAN YL， ZHAO LL， LI JH， et al. Information technology of modern NIR spectral analysis[J]. Spectroscopy and Spectral Analysis， 2000， 20（6）： 777-780.
[13] DELWICHE SR， HRUSCHKA WR. Protein content of bulk wheat from near??infrared reflectance of individual kernels [J]. Cereal Chemistry， 2000， 77（1）： 86-88.
[14] YAN YL， ZHAO LL， HAN DH， et al. Near infrared spectroscopic analysis basis and application[M]. Beijing： China Light Industry Press， 2005： 3-10.
[15] YAN YL， CHEN B， ZHU DZ， et al. Principle， technique and application of near infrared spectroscopy[M]. Beijing： China Light Industry Press， 2013： 1-7.
[16] XU GT， YUAN HF， LU WZ. General situation and progress of near??infrared spectroscopy analysis instrument[J]. Modern Scientific Instruments， 1997（3） ： 9-11.