近红外漫透射光补偿法无损快速检测大米直链淀粉
|
摘要
针对大米长波近红外漫透射光谱噪声大的问题,自行搭建了3种光谱采集系统用于分析波长范围为900～1700 nm的大米漫透射光补偿,采集了62个样本的大米红外光谱曲线,并进行了归一化、SG平滑、Savitzky-Golay卷积求导预处理,用偏最小二乘回归法对大米直链淀粉含量进行了建模分析,比较分析同种大米在不同厚度下的光补偿前后漫透射光谱曲线,对比漫反射、漫透射、漫透射光补偿结果,并对光补偿前后的结果进行了显著性分析。结果表明,光补偿前,随着样品厚度增加,大米直链淀粉含量预测模型结果先变好,但是随着样品厚度进一步增加,透射光强随之变弱,噪声变大,模型建模效果变差。样品厚度为9 mm时,大米近红外漫透射直链淀粉预测模型效果最好,校正集相关系数(R_C)为0.9103,校正集均方根误差(RMSEC)为1.4209%;预测集相关系数(R_P)为0.9049,预测集均方根误差(RMSEP)为1.5654%;光补偿后,大米近红外漫透射光补偿光谱曲线噪声显著改善,特别是经预处理后光谱曲线噪声在1203和1465 nm附近的光谱吸收处改善明显,并且不同样品厚度条件下的预测模型精度均有显著提高。大米样品厚度为9 mm时,直链淀粉光补偿预测模型效果最佳,模型校正集相关系数(R_C)提升到0.9654,校正集均方根误差(RMSEC)降低到0.8902%;预测集相关系数(R_P)达到0.9577,预测集均方根误差(RMSEP)降低到1.4261%,并且光补偿后的显著性较光补偿前有所降低,与相关研究相比,模型的相关系数和误差均有所改善。最后,选用没有参与建模的20个样品对光补偿模型进行了外部检验,模型相关系数为0.9363,均方根误差为1.4139%,RPD为2.85。结果表明,光补偿方法可以有效解决大米长波近红外因穿透力相对较弱而引起光谱噪声大的问题,提高大米直链淀粉预测模型的精度,可以实现颗粒大米直链淀粉含量的快速无损检测,为大米品质检测分级提供技术支撑。
To solve the problem of low signal-to-noise ratio in detection of rice amylose content by long wavelength near infrared diffuse transmission spectroscopy, 3 spectral data acquisition systems were built. In the wavelength range of 900-1700 nm, the original spectrum curves of 62 kinds of rice samples before and after light compensation were collected, normalized, smoothed by SG, and treated by Savitzky-Golay derivative convolution. After that, partial least squares regression modeling was performed for amylose content analysis of rice, by which the light diffuse transmission spectrum curves of rice sample with different thicknesses before and after light compensation were comparatively investigated, and the physical and chemical standard values before and after light compensation were subjected to significant analysis. The results showed that the prediction model was improved with the increase of sample thickness before light compensation, but with the increase of sample thickness, the transmitted light intensity was weakened, the signal-to-noise ratio was decreased, the noise was increased, and the model modeling effect became worse. When the sample thickness was 9 mm, the prediction model of near-infrared diffuse transmission spectroscopy for rice amylose had the best result. Under this condition, the correlation coefficients of correction set(R_C) and prediction set(R_P) were 0.9103 and 0.9049, respectively, and their root mean square errors(RMSEC and RMSEP) were 1.4209% and 1.5654%, respectively. After the optical compensation, the signal-to-noise ratio of near-infrared diffuse transmission spectrum curve of the rice sample was significantly improved, especially the spectral absorption near 1203 nm and 1465 nm after the pretreatment, and the prediction model accuracy under different sample thicknesses was also significantly improved. When the thickness of rice samples was 9 mm, the effect of optical compensation prediction model for amylose detection was the best. The correlation coefficient of model correction set(R_C) and prediction set(R_P) were increased to 0.9654 and 0.9577, respectively, and their root mean square errors(RMSEC and RMSEP) were reduced to 0.8902% and 1.4261%, respectively. In addition, the significance of the light compensation was reduced compared with that before the light compensation, and the correlation coefficient and error of the model were improved compared with other methods. Finally, 20 samples were selected for external testing of the optical compensation model with a sample thickness of 9 mm, the model correlation coefficient(RV) was 0.9363, the root mean square error(RMSEV) was 1.4139%, and the RPD value was 2.85. The results showed that the light compensation method could effectively solve the problem of low signal-to-noise ratio caused by relatively weak penetrating power of diffuse transmission spectrum in detection of rice amylose by long wavelength near infrared spectrometry, which improved the accuracy of forecasting model, realized the rapid and nondestructive testing of rice amylose content, and provided technical support for rice quality detection.
To solve the problem of low signal-to-noise ratio in detection of rice amylose content by long wavelength near infrared diffuse transmission spectroscopy, 3 spectral data acquisition systems were built. In the wavelength range of 900-1700 nm, the original spectrum curves of 62 kinds of rice samples before and after light compensation were collected, normalized, smoothed by SG, and treated by Savitzky-Golay derivative convolution. After that, partial least squares regression modeling was performed for amylose content analysis of rice, by which the light diffuse transmission spectrum curves of rice sample with different thicknesses before and after light compensation were comparatively investigated, and the physical and chemical standard values before and after light compensation were subjected to significant analysis. The results showed that the prediction model was improved with the increase of sample thickness before light compensation, but with the increase of sample thickness, the transmitted light intensity was weakened, the signal-to-noise ratio was decreased, the noise was increased, and the model modeling effect became worse. When the sample thickness was 9 mm, the prediction model of near-infrared diffuse transmission spectroscopy for rice amylose had the best result. Under this condition, the correlation coefficients of correction set(R_C) and prediction set(R_P) were 0.9103 and 0.9049, respectively, and their root mean square errors(RMSEC and RMSEP) were 1.4209% and 1.5654%, respectively. After the optical compensation, the signal-to-noise ratio of near-infrared diffuse transmission spectrum curve of the rice sample was significantly improved, especially the spectral absorption near 1203 nm and 1465 nm after the pretreatment, and the prediction model accuracy under different sample thicknesses was also significantly improved. When the thickness of rice samples was 9 mm, the effect of optical compensation prediction model for amylose detection was the best. The correlation coefficient of model correction set(R_C) and prediction set(R_P) were increased to 0.9654 and 0.9577, respectively, and their root mean square errors(RMSEC and RMSEP) were reduced to 0.8902% and 1.4261%, respectively. In addition, the significance of the light compensation was reduced compared with that before the light compensation, and the correlation coefficient and error of the model were improved compared with other methods. Finally, 20 samples were selected for external testing of the optical compensation model with a sample thickness of 9 mm, the model correlation coefficient(RV) was 0.9363, the root mean square error(RMSEV) was 1.4139%, and the RPD value was 2.85. The results showed that the light compensation method could effectively solve the problem of low signal-to-noise ratio caused by relatively weak penetrating power of diffuse transmission spectrum in detection of rice amylose by long wavelength near infrared spectrometry, which improved the accuracy of forecasting model, realized the rapid and nondestructive testing of rice amylose content, and provided technical support for rice quality detection.
引文
1 Li H Y,Prakash S,Nicholson T M,Fitzgerald M A,Gilbert R G.Food Chem.,2016,196:702-711
2 Qin K X,Liu L L,Liu T Y,Cheng W H,Shi Y G.J.Northeast Agric.University (English Edition),2014,21(3):60-67
3 XIA Fan,DONG Yue,ZHU Lei,ZHANG Ai-Jing,WANG Peng-Jie,YUAN Jian,GAO Yu-Long.Grain Science and Technology and Economy,2018,43(5):100-107夏凡,董月,朱蕾,张爱静,王鹏杰,袁建,高瑀珑.粮食深加工及食品,2018,43(5):100-107
4 Bagchi T B,Sharma S,Chattopadhyay K.Food Chem.,2016,191:21-27
5 LIU Shan,WANG Jian-Jun,FAN Xiao-Juan,WANG Lin-You,ZHANG Li-Xia.Journal of the Chinese Cereals and Oils Association,2015,30(3):140-146刘姗,王建军,范小娟,王林友,张礼霞.中国粮油学报,2015,30(3):140-146
6 Park C E,Kim Y S,Park K J,Kim B K.J.Stored Products Res.,2012,48(48):25-29
7 Meullenet J F,Mauromoustakos A,Horner T B,Marks B P.Cereal Chem.,2002,79(1):52-57
8 SHU Qing-Yao,WU Dian-Xing,XIA Ying-Wu,GAO Ming-Wei,Mclung A.Chinese Journal of Rice Science,1999,(3):189-192舒庆尧,吴殿星,夏英武,高明尉,Anna McClung.中国水稻科学,1999,(3):189-192
9 Delwiche S R,Mckenzie K S,Webb B D.Cereal Chem.,1996,73(2):257-263
10 Sampaio P S,Soares A,Castanho A,Almeida A C,Oliveira J,Brites C.Food Chem.,2018,242:196-204
11 Wu J G,Shi C H.Field Crops Res.,2004,87(1):13-21
12 Siriphollakul P,Nakano K,Kanlayanarat S,Ohashi S,Sakai R,Rittiron R,Maniwara P.LWT Food Sci.Technol.,2017,79:70-77
13 SHU Qing-Yao,WU Dian-Xing,XIA Ying-Wu,GAO Ming-Wei.Scientia Agricultura Sinica,1999,(4):94-99舒庆尧,吴殿星,夏英武,高明尉.中国农业科学,1999,(4):94-99
14 CHENG Feng,SHU Gong-Chen,ZHANG Hong-Rui,ZHANG Shi-Yong,SHUN Ya-Wei,ZHANG Zheng-Qiu.China Rice,2009,(2):38-39陈峰,孙公臣,张洪瑞,张士永,孙亚伟,张正球.中国稻米,2009,(2):38-39
15 LIU Jian-Xue,WU Shou-Yi,FANG Ru-Ming.Transactions of the Chinese Society of Agricultural Machinery,2001,(2):55-57刘建学,吴守一,方如明.农业机械学报,2001,(2):55-57
16 Kawamura S,Natsuga M,Takekura K,Itoh K.Comput.Electronics Agric.,2003,39(1):115-126
17 Kim H,Kim O,Kwak H S,Kim S S,Lee H.J.Sens.Studies,2017,32(4):e12273
18 WANG Fan,LI Yong-Yu,PENG Yan-Kun,ZHENG Xiao-Chun.Spectroscopy and Spectral Analysis,2016,36(10):3185-3189王凡,李永玉,彭彦昆,郑晓春.光谱学与光谱分析,2016,36(10):3185-3189
19 LIU Yan-De,WU Ming-Ming,LI Yi-Fan,SHUN Xu-Dong,HAO Yong.Spectroscopy and Spectral Analysis,2017,37(8):2424-2429刘燕德,吴明明,李轶凡,孙旭东,郝勇.光谱学与光谱分析,2017,37(8):2424-2429
20 WANG Fan,LI Yong-Yu,PENG Yan-Kun,SUN Hong-Wei,LI Long.Chinese J.Anal.Chem.,2018,46(9):1424-1431王凡,李永玉,彭彦昆,孙宏伟,李龙.分析化学,2018,46(9):1424-1431
21 TANG Xing-Tian,WU Wen-Jin,GUO Ting-Ting,JIA Shi-Qiang,YAN Yan-Lu,AN Dong.Spectroscopy and Spectral Analysis,2012,32(6):1531-1534唐兴田,邬文锦,郭婷婷,贾仕强,严衍禄,安冬.光谱学与光谱分析,2012,32(6):1531-1534
22 ISO 6647-2:2015 Rice-Determination of Amylose Content-Part 2:Routine Method
23 PENG Yan-Kun.Nondestructive Optical Technology For Agro-Food Quality and Safety Assessment.Beijing:Science Press,2016:61彭彦昆.农畜产品品质安全光学无损快速检测技术.北京:科学出版社,2016:61
2 Qin K X,Liu L L,Liu T Y,Cheng W H,Shi Y G.J.Northeast Agric.University (English Edition),2014,21(3):60-67
3 XIA Fan,DONG Yue,ZHU Lei,ZHANG Ai-Jing,WANG Peng-Jie,YUAN Jian,GAO Yu-Long.Grain Science and Technology and Economy,2018,43(5):100-107夏凡,董月,朱蕾,张爱静,王鹏杰,袁建,高瑀珑.粮食深加工及食品,2018,43(5):100-107
4 Bagchi T B,Sharma S,Chattopadhyay K.Food Chem.,2016,191:21-27
5 LIU Shan,WANG Jian-Jun,FAN Xiao-Juan,WANG Lin-You,ZHANG Li-Xia.Journal of the Chinese Cereals and Oils Association,2015,30(3):140-146刘姗,王建军,范小娟,王林友,张礼霞.中国粮油学报,2015,30(3):140-146
6 Park C E,Kim Y S,Park K J,Kim B K.J.Stored Products Res.,2012,48(48):25-29
7 Meullenet J F,Mauromoustakos A,Horner T B,Marks B P.Cereal Chem.,2002,79(1):52-57
8 SHU Qing-Yao,WU Dian-Xing,XIA Ying-Wu,GAO Ming-Wei,Mclung A.Chinese Journal of Rice Science,1999,(3):189-192舒庆尧,吴殿星,夏英武,高明尉,Anna McClung.中国水稻科学,1999,(3):189-192
9 Delwiche S R,Mckenzie K S,Webb B D.Cereal Chem.,1996,73(2):257-263
10 Sampaio P S,Soares A,Castanho A,Almeida A C,Oliveira J,Brites C.Food Chem.,2018,242:196-204
11 Wu J G,Shi C H.Field Crops Res.,2004,87(1):13-21
12 Siriphollakul P,Nakano K,Kanlayanarat S,Ohashi S,Sakai R,Rittiron R,Maniwara P.LWT Food Sci.Technol.,2017,79:70-77
13 SHU Qing-Yao,WU Dian-Xing,XIA Ying-Wu,GAO Ming-Wei.Scientia Agricultura Sinica,1999,(4):94-99舒庆尧,吴殿星,夏英武,高明尉.中国农业科学,1999,(4):94-99
14 CHENG Feng,SHU Gong-Chen,ZHANG Hong-Rui,ZHANG Shi-Yong,SHUN Ya-Wei,ZHANG Zheng-Qiu.China Rice,2009,(2):38-39陈峰,孙公臣,张洪瑞,张士永,孙亚伟,张正球.中国稻米,2009,(2):38-39
15 LIU Jian-Xue,WU Shou-Yi,FANG Ru-Ming.Transactions of the Chinese Society of Agricultural Machinery,2001,(2):55-57刘建学,吴守一,方如明.农业机械学报,2001,(2):55-57
16 Kawamura S,Natsuga M,Takekura K,Itoh K.Comput.Electronics Agric.,2003,39(1):115-126
17 Kim H,Kim O,Kwak H S,Kim S S,Lee H.J.Sens.Studies,2017,32(4):e12273
18 WANG Fan,LI Yong-Yu,PENG Yan-Kun,ZHENG Xiao-Chun.Spectroscopy and Spectral Analysis,2016,36(10):3185-3189王凡,李永玉,彭彦昆,郑晓春.光谱学与光谱分析,2016,36(10):3185-3189
19 LIU Yan-De,WU Ming-Ming,LI Yi-Fan,SHUN Xu-Dong,HAO Yong.Spectroscopy and Spectral Analysis,2017,37(8):2424-2429刘燕德,吴明明,李轶凡,孙旭东,郝勇.光谱学与光谱分析,2017,37(8):2424-2429
20 WANG Fan,LI Yong-Yu,PENG Yan-Kun,SUN Hong-Wei,LI Long.Chinese J.Anal.Chem.,2018,46(9):1424-1431王凡,李永玉,彭彦昆,孙宏伟,李龙.分析化学,2018,46(9):1424-1431
21 TANG Xing-Tian,WU Wen-Jin,GUO Ting-Ting,JIA Shi-Qiang,YAN Yan-Lu,AN Dong.Spectroscopy and Spectral Analysis,2012,32(6):1531-1534唐兴田,邬文锦,郭婷婷,贾仕强,严衍禄,安冬.光谱学与光谱分析,2012,32(6):1531-1534
22 ISO 6647-2:2015 Rice-Determination of Amylose Content-Part 2:Routine Method
23 PENG Yan-Kun.Nondestructive Optical Technology For Agro-Food Quality and Safety Assessment.Beijing:Science Press,2016:61彭彦昆.农畜产品品质安全光学无损快速检测技术.北京:科学出版社,2016:61