Determination of oil, protein and moisture content in whole seeds of three oleiferous Brassica species using near-infrared reflectance spectroscopy

Abstract

The near infrared reflectance spectroscopy (NIRS) was used for non destructive estimation of
biochemical components in seed for three Brassica species, irrespective of harvest season and seed coat
colour heterogeneity of the samples. Self – pollinated seeds of 132 yellow or brown seeded genotypes of B.
juncea, B. napus and B. rapa were analysed through wet chemistry and NIRS scanning. The nuclear
magnetic resonance (NMR), oven drying, and Kjeldahl nitrogen estimation were used as reference methods
for oil, moisture and protein content, respectively. Considerable variations were observed between and within
the species for these constituents. The original Log 1/R seed spectrums showed high absorption in the range
of 1124- 1300, 1592- 1868, 1920- 2104, and 2208- 2356 nm wavelength. A considerable base line and
absorption intensity variations were observed for between and within the species respectively. The 2nd
derivative transformation with gap = 2, segment size = 2 nm was established as most appropriate for
calibration development. The best combined species calibrations resulted in the best calibration statistics for
the estimation of the oil (SECV = 1.30, R2 = 0.94, SEC = 1.18), moisture (SECV = 0.12, R2 = 0.87, SEC =
0.39) and protein content (SECV = 12.19, R2 = 0.91, SEC = 2.18). A high linear correlation (r2 > 0.85) and
non – significant difference (P > 0.05) was observed between the NIRS and the reference method
estimations, proving the robustness of the calibration equations.