| Rice is one of the key crops, which is high yields, and distributes in most area of China. Improving the quality of rice is an important task to meet the increase of people's living situation. Amylose content (AC) and protein content(PC) are primary factors which influence the traits of cooking and nutrition quality. During the breeding procedure, lots of derived lines with good quality were selected from various candidate varieties, and breeders usually expect to get superior vigor seeds which could be selected from the low generations. While in the realistic breeding, the quality of derived lines used to be tested with chemical methods, by which the vigor of individuals tested was thoroughly destroyed. Besides that, those methods always cost much more. In order to enhance the choose efficiency, such methods with speediness, high efficiency, and simpleness but without destroying the vigor of the tested materials are urgently expected by current breed experts.In recent years, near infrared(NIR) spectroscopy has greatly developed in analysis and measuration of quality, which provides a new and efficiency means for current breeding by speediness, high efficiency, and simpleness but without destroying the tested materials. So that it is significant to reduce breeding scales, shorten times and enhance breeding efficiency. To make use of the advantage of NIR spectroscopy, NITS should be applied to establish corresponding models for different samples, analysis contents and areas. Combined with NITS and chemical methods, this paper has established eight NITS models of AC and PC in rice, which is suitable for Chongqing.Firstly, 612 samples of rice, including hybride rice and conventional rice for calibration and exterior-validation were collected and obtained the NITS scan spectrums by the Tecator Infratec 1255 Food and Feed Analyzer.Secondly, AC and PC of samples were mensurated by chemical procedure. The results showed that AC ranged from 2.011% to 27.763% and PC from 7.270% to 15.434% respectively. That could be applicable and representative to satisfy the rule rice's test.Thirdly, combined with NITS spectrum and the result of chemical analysis, the NITS spectrums were pretreated by the methods of Detrending, Standard Normal Variate, Multiplicative Scatter Correction. While Comparing with the regression methods of PCR, PLS, and MPLS, this paper chose SNV and Trending D, mathematics using l,4,4,l,and the regressive analysis selecting MPLS.Finally, this paper has established eight calibration models of brown and milled in conventional and hybride rice separately, including respective AC and PC models, while evaluated the predication capacity of the calibration equations were evaluated by cross-validation and external validation. The results indicated that the regression squared (RSQ) of the calibration in AC models of hybrid brown rice and milled rice were 0.873 and 0.922, in PC models 0.963 and 0.978, the standard errors for calibration (SEC) in AC models of hybrid brown rice and milled rice were 1.100 and 0.956, in PC models 0.202 and 0.155, while conventional rice were 0.252 and 0.247, the RSQ of conventional were 0.924,0.939,0.953 and 0.946 respectively , the SEC of conventional rice were 1.537,1.547,0.252 and 0.247 ,the RSQ of cross-validation in AC models were 0.866, 0.901, 0.892 and 0.921, in PC models were 0.202., 0.155, 0.252, 0.247 respectively, and the RSQ of exterior-validation in AC models are 0.951, 0.935, 0.912 and 0.918,in PC models are 0.960, 0.949, 0.969 and 0.956 respectively. The results showed that there be significant correlation and lower error between near-infrared value and true value in all models. On the other hand the models of conventional rice could be applied in early generation selecting with speediness, high efficiency, and simpleness but without destroying the breeding tested materials, while the models of hybride could be used in AC and PC of new compages's identification, so as to promote rice quality improvement, and enhance breeding efficiency. |
PDF Full Text Request