抄録
The feasibility of near-infrared diffuse reflectance spectroscopic analysis (NIR analysis) for estimating the ratio of true seed weight to fruit weight (T/F) as well as water content was examined in sugar beet (Beta vulgaris L.) seeds of 61 F1 hybrid lines and 4 standard cultivars. For the calibration, partial least squares (PLS) regression was carried out with second derivative spectra and the measured data using attached software (NSAS). For estimating T/F, calibration using 7 factors was the most valid with a correlation coefficient of calibration (R) of 0.943, standard error of calibration (SEC) of 1.26% and standard error of prediction (SEP) of 1.40%. By this calibration, sugar beet varieties could be classified into 4–5 levels according to the estimated T/F. For the estimation of the water content of sugar beet seed, calibration using 14 factors was optimal. The calibration was highly accurate since the R, SEC and SEP was 0.999, 0.23% and 0.27% respectively. Consequently, the true seed weight of sugar beet could be nondestructively and rapidly estimated by NIR analysis and weighing the air-dried seed. This technique should be useful in breeding selection for higher true seed weights, which would thereby improve the early growth of sugar beet varieties.
