科研之友
微信
新浪微博
Facebook
分享链接
摘要
For the development of calibration equations, optical and chemical database, including winter and spring wheat (Triticum aestivum L.) samples, was collected over the period 1995-2008. Reflectance spectra were recorded in the range of 400-2500 nm by NIRS-6500. Moisture, protein and starch were determined by reference methods. The effects of calibrated optical data math treatment. scatter correction, calibrated spectrum wavelength range. regression method on the accuracy of equations being developed were estimated. Depending oil calibration model, equations of different accuracy in calibration were developed: for moisture values, RSQ = 0.817-0.941, SEC = 0.366-0.646. for protein content determination, RSQ = 0.716-0.947, SEC = 0.442-1.009, for starch content, RSQ = 0.838-0.869 SEC = 1.152-1.302.
It was determined that adjustment of different parameters of calibration is an effective means for the improvement of quality prediction by NIRS-6500 spectroscopy. Equations having the best statistical parameters in prediction of control sample set for each quality component were selected. Calibration of spectra segment 1100-2500 nm, treated by first derivative was the most useful for moisture calibration. Optimal calibration model for protein content assessment was as follows: wavelength range 400-2460, 600-2460 or 700-2460 nm, math treatment 2, 4, 4, 1, scatter correction NSVD and modified partial least squares (MPLS) regression. The most robust equations for starch evaluation were developed when for NIR spectra segment (1100-2500 nm) or larger wavelength range (500-2500 nm) the first derivative was applied. The equations developed for wheat grain moisture assessment and protein content are characterised by high precision. The equations developed for the determination of starch concentration in grain can be estimated as giving good or satisfactory precision.
- 出版日期2009
