论文部分内容阅读
草莓是重要的草本农作物,固体可溶物的含量是影响草莓品质的主要目标之一。采用傅里叶变换近红外光谱(FT-NIRS)结合统计计量学方法对草莓的固体可溶物含量进行定量检测。基于偏最小二乘回归(PLSR)算法,通过将光谱矩阵X和化学浓度矩阵Y进行正交分解,获取对应的综合变量(得分矩阵),结合采用多元散射校正(MSC)预处理方法,提取草莓固体可溶物的近红外光谱信息响应,讨论和比较MSC方法作用于PLSR变量分解之前或者之后的近红外分析模型预测效果。针对变量分解之后再进行光散射校正的模型(PLS+MSC+R)具有相对较高的模型效度和信度,分别利用均方根偏差(RMSEP)和相对偏差(RSD)来量化,RMSEP=0.367(%),RSD=3.51%。结果表明,MSC光散射校正技术和PLSR回归方法的综合运用能够提升草莓固体可溶物的近红外光谱建模预测能力,如果进一步结合现代食品工业的在线检测技术,有利于优质草莓品种的培育。
Strawberry is an important herbaceous crop. The content of solid solubles is one of the main goals of strawberry quality. The FT-NIRS and statistical metrology methods were used to quantitatively detect the soluble solids in strawberry. Based on Partial Least Squares Regression (PLSR) algorithm, the corresponding comprehensive variables (scoring matrix) were obtained by orthogonally decomposing spectral matrix X and chemical concentration matrix Y. Strawberry was extracted by using the method of multivariate scatter correction (MSC) NIRS information response of solid solubles was discussed and compared with that predicted by MSC method before and after decomposition of PLSR variables. The model (PLS + MSC + R), which performs light scattering correction after variable decomposition, has relatively high model validity and reliability, and is quantified using root mean square deviation (RMSEP) and relative deviation (RSD) respectively. RMSEP = 0.367 (%), RSD = 3.51%. The results showed that the combined application of MSC light scattering correction and PLSR regression could enhance the ability of NIRS modeling and prediction of strawberry solid solubles. Combining the on - line detection technology of modern food industry, it is beneficial to cultivate high quality strawberry varieties.