NIR Spectroscopy Model Construction And Application In The Research Of Crop Seed Qualities

In recent years, near infrared (NIR) spectroscopy has made rapid progress as analytical technique. Being capable of making nondestructive, rapid, high efficient and convenient analysis, NIR technique is suitable to analyze seed quality traits.The accuracy, reliability, and application scope of NIR method are depended on the calibration model developed by the database based on routine analytical methods. The calibration model of NIR possesses sample and constituent dependence when is used for predication. To make use of the advantage of NIR spectroscopy, the individual calibration models must be developed according to the corresponding sample status and constituents. Thus, the key work of this technique is to develop the robust and accurate calibration models.The goal of this project was to develop calibration models for seed quality traits based on the characteristics of the NIR instrument, which could be capable of analyzing amount of seed samples, and to reduce the analysis cost greatly and to increase the analysis efficiency and to accelerate the usage frequency of the instrument. Meanwhile, the calibration models for small sample and single grain were constructed according to the requirement of crop breeding and the analytical scope of the NIR instrument, which needed to be enlarged and exploited.The seed samples of crops such as rice, rapeseed, barley and maize were chosen as materials and collected in large scale and various specific seed samples were created. Many important constituents of seed sample were determined and corresponding NIR spectra of various samples were collected, therefore, the original databases for NIR calibration were created. According to the requirement of quality traits studies, thecorresponding calibration models were developed by using various chemometrics methods.The main results were as follows: The NIR calibration models, which possess an accuracy and reliable predication, were constructed for rice common constituents such as amylose content, gel consistency, alkali spreading value and RVA. The further calibration models for rice nutrient quality traits like protein content and various amino acid contents were developed and the scope of the NIR application on rice nutrient assay was enlarged. For predicating the samples used in the genetic and breeding purpose, the models, which had possessed the information including a large database, small sample size and intact seed samples, were created for rapeseed oil, protein, glucosinolate and fatty acids contents. The NIR calibration models for barley protein and glucan contents were developed, simultaneously and created the models which can be used to predicate maize moisture and protein content precisely. When developing the various models, the parameter and selecting methods of calibration set were investigated and chemometrics parameter for models were optimized in order to make a base for constructing future normative analysis methods.