论文部分内容阅读
近红外光谱学是近十年来发展最快、最引人注目的光谱分析技术之一,其高精度测量依赖于化学计量学方法以准确提取光谱信息.针对近红外光谱单尺度传统建模方法中存在的信息易丢失问题,发展了一种多尺度建模新方法.多尺度建模可有效协同利用信号的时/频多尺度特性,并将多尺度特性以加权形式统一映射到多元校正空间,有效避免了信息丢失.该算法成功地应用于面粉中掺杂有毒非法添加剂硼砂含量检测,经验证后模型的预测值与真实值的相关系数和预测均方根误差分别为0.974和0.0019,其预测相对误差为1.9%.研究结果表明,多尺度建模方法完全满足近红外光谱高精度测量的要求.
Near-infrared spectroscopy is one of the fastest growing and most spectacular spectroscopic techniques in recent ten years, and its high-precision measurement relies on chemometric methods to accurately extract spectral information.For the single-scale conventional modeling of near-infrared spectroscopy The existing information is easy to be lost, and a new multi-scale modeling method is developed.Multiscale modeling can effectively utilize the time / frequency multi-scale features of signals and map the multi-scale features to the multivariate correction space in a weighted form, Which effectively avoids the loss of information.The algorithm was successfully applied to the detection of borax with toxic additives in flour.The correlation coefficient and predicted root mean square error (RMSE) between the predicted and the true values of the model were 0.974 and 0.0019 respectively, and its prediction The relative error is 1.9% .The research results show that the multi-scale modeling method completely meets the requirements of high-precision NIRS measurement.