Models of Oil and Protein Content by NIRS Analysis in Brassica Oleracea Var. Acephala

田锦兰,唐章林,徐新福,王瑞,李加纳

DOI: https://doi.org/10.13430/j.cnki.jpgr.2009.03.023

2009-01-01

Abstract:Spectroscopy data for seeds of 807 accessions of Brassica oleracea var.acephala were collected by near infrared reflectance spectroscopy(NIRS) and contents of seed oil and meal protein of 250 accessions selected from 807 accessions were determined by classical chemical methods.Mathematical models of NIRS analysis were established using partial least square(PLS),modified partial least squares(MPLS) and different technologies of wavelengths and mathematical treatments.The correlation coefficients(1-VR) of internal cross validation test of the models approximated to 1,the standard errors of cross-validation(SECV) and the standard errors of calibration(SEC) were small.In external test of the models,the determination coefficients of validation(RSQ) were all above 0.9,the relative mean deviations were below 4%,the corrected standard errors of performance(SEP) and average bias between predictive value(BIAS) were much smaller.So,the two predictive models had high precision and can be used in quality breeding programs.