摘要
【目的】应用近红外光谱漫反射技术在线检测脐橙内部的可溶性固形物含量（SSC）。【方法】以0.3m/s的速度、400W的光照强度获取脐橙（脐橙样品为97个,其中74个为校正集,23个样品为预测集）的漫反射光谱;对比不同光谱预处理方法（平滑、一阶微分、二阶微分等）对偏最小二乘回归（PLSR）所建预测模型性能的影响,建立PLSR、主成分回归（PCR）和多元线性回归（MLR）在线检测脐橙可溶性固形物含量的预测模型。【结果】在520~1 000nm光谱范围,卷积平滑（S-G）能有效提高光谱的信噪比,改善模型预测精度;基于PLSR所建立的预测模型较PCR和MLR更为理想,其预测相关系数（RP）为0.90,预测均方根误差（RMSEP）为0.61。【结论】利用在线近红外光谱技术检测脐橙可溶性固形物含量是可行的。
Abstract： [Objective] The research focused on online detection of soluble solids content （SSC） in na- vel oranges using near-infrared spectroscopy. [Method] Navel oranges spectrum was collected using dif- fuse reflectance with movement speed of 0.3 m/s and light intensity of 400 W. A total of 97 navel oranges were measured in the experiment,among which 74 were within the calibration set and 23 were within vali- dation set. Different pretreatment methods such as Savitzky-Golay smooth, first derivative, second deriva- tive and so on were compared. Different calibration models were developed based on partial least squares （PLSR） ,principal component regression （PCR）,and multiple-linear regression （MLR）. [Result] In spec- tral range of 520--1 000 nm,the S-G smooth could increase S/N ratio and improve the performance of mod- els. The best calibration model was based on PLSR method with the prediction correlation coefficients （Re） of 0.90 and the root mean square errors of prediction （RMSEP） of 0.61. [Conclusion] The established on- line detection of SSC in navel oranges is feasible.
出处
《西北农林科技大学学报（自然科学版）》
CSCD
北大核心
2014年第3期186-190,共5页
Journal of Northwest A&F University(Natural Science Edition)
基金
国家"863"高技术研究发展计划项目(2012AA101906)
科技部农业科技成果转化项目(2011GB2C500008)
赣鄱英才555工程领军人才培养计划(2011-64)
江西省光电检测工程技术研究中心资助项目(赣科发财字[2012]155号)
江西省研究生创新专项资金项目(YC2013-S166)
关键词
脐橙
近红外光谱
在线检测
可溶性固形物
navel orange
near-infrared diffuse spectroscopy
online detection
soluble solids content
作者简介
刘燕德（1967-），女，江西泰和人，教授，博士生导师，主要从事光机电技术及应用研究。E—mail：jxliuyd@163．com
