基于近红外光谱的马铃薯匀浆常规成分快速分析模型的构建

打开文本图片集

中图分类号：S127 文献标识号：A 文章编号：1001-4942（2025）07-0159-07

AbstractIn this study，the near-infrared prediction models for detecting five conventional quality indexes including protein，dry matter，starch，reducing sugar and vitamin C of fresh potato homogenate were established based on near-infrared reflectance spectroscopy（NIRS）and chemical test method，in order to provide a theoretical basis for rapid detection of potato nutritional quality.In the all fresh potato homogenate samples， 532 ones were selected to form the calibration set and 135ones were used to form the validations set.The NIRS DS2500 spectrometer was used for spectrum scanning，and the five quality indexes were determined using the chemical methods in state and industrial standards.The prediction models of the five nutritional indexes were established using the modified partial least squares（MPLS），and their pros and cons were verified. The results showed that the cross-validation correlation coefficients （1-VR）of protein，dry matter，starch，reducing sugar and vitamin C contents were 0.837，0.961， 0.907，0.941 and 0.969，all exceeding 0.80，so the established models could be used for actual prediction，among which，those for dry matter，starch，reducing sugar and vitamin C could be used for accurate prediction.

KeyWordsNear-infrared reflectance spectroscopy （NIRS）；Potato homogenate； Nutritional quality;Modified partial least squares ； Calibration model

马铃薯（SolanumtuberosumL.）不仅是常见的蔬菜，更是四大主粮之一[1]。（剩余9722字）