Premium
Application of near‐infrared reflectance spectroscopy to estimate the bulk density of Ethiopian mustard seeds
Author(s) -
Velasco Leonardo,
FernándezMartínez José M,
De Haro Antonio
Publication year - 1998
Publication title -
journal of the science of food and agriculture
Language(s) - English
Resource type - Journals
SCImago Journal Rank - 0.782
H-Index - 142
eISSN - 1097-0010
pISSN - 0022-5142
DOI - 10.1002/(sici)1097-0010(199807)77:3<312::aid-jsfa39>3.0.co;2-z
Subject(s) - calibration , partial least squares regression , linear regression , coefficient of determination , mathematics , regression analysis , near infrared spectroscopy , statistics , analytical chemistry (journal) , chemistry , physics , chromatography , optics
This study was conducted to test the applicability of NIRS to estimate the bulk density of Ethiopian mustard seeds. A set of samples from the 1992 crop with bulk density values between 262 and 676 kg m −3 was used to develop a calibration equation for this trait. The best equation was obtained by using the original log (1/ R ) spectra, without any mathematical pretreatment, and modified partial least squares (MPLS) regression. The coefficient of determination in external validation ( r 2 ) was 0·92, which indicated a high reliability in the estimation of bulk density by NIRS. This equation was successively expanded with new samples from the 1993 and 1994 crops. The calibration equation developed in 1994, including 356 samples from the 1992, 1993 and 1994 crops, was validated with a total of 362 samples from the 1995 crop, ranging from 204 to 697 kg m −3 . The r 2 was 0·92, indicating the high robustness of the 3‐year calibration equation to estimate the bulk density of samples from populations not represented in the calibration set. A 3‐year calibration was also developed by multiple linear regression (MLR) and a restriction to 19 wavelengths available in common filter instruments, resulting in an r 2 =0·88 in the validation with samples from 1995, which suggests that this trait may be also accurately estimated in filter instruments. This study demonstrated that the bulk density of Ethiopian mustard seeds can be easily and accurately estimated by NIRS, which facilitates the rapid identification of undesirable low bulk density samples in routine analysis. © 1998 SCI.