光谱学与光谱分析, 2013, 33 (10): 2843, 网络出版: 2013-10-23
以高光谱数据有效预测苹果可溶性固形物含量
Effectively Predicting Soluble Solids Content in Apple Based on Hyperspectral Imaging
高光谱成像 苹果 可溶性固形物含量 变量选择 多元校正分析 Hyperspectral imaging Apple Soluble solids content Variable selection Multivariate calibration analysis
摘要
从高光谱数据中选取能够有效进行内部品质检测的特征波长, 是利用高光谱成像技术进行水果品质定量分析的关键。 本文采用遗传算法(GA)、 连续投影算法(SPA)和GA-SPA算法分别从400~1 000 nm的苹果高光谱图像中提取特征波长, 利用偏最小二乘法(PLS)、 最小二乘支撑向量机(LS-SVM)和多元线性回归(MLR)建模进行苹果可溶性固形物含量(SSC)的定量分析并进行了综合比较。 160个样品中, 120个用于建模, 40个用于预测。 比较发现SPA-MLR模型获得了最好的结果, R2p, RMSEP和RPD分别为0.950 1, 0.308 7和4.476 6。 结果表明: SPA能够有效地用于高光谱数据的变量选择, 利用SPA-MLR可建立稳健的苹果SSC预测模型, 较少的有效变量和MLR模型的易解释性表明该模型在在线检测和便携式仪器开发中具有较大的应用潜力。
Abstract
It is very important to extract effective wavelengths for quantitative analysis of fruit internal quality based on hyperspectral imaging. In the present study, genetic algorithm (GA), successive projections algorithm (SPA) and GA-SPA combining algorithm were used for extracting effective wavelengths from 400~1 000 nm hyperspectral images of Yantai “Fuji” apples, respectively. Based on the effective wavelengths selected by GA, SPA and GA-SPA, different models were built and compared for predicting soluble solids content (SSC) of apple using partial least squares (PLS), least squared support vector machine (LS-SVM) and multiple linear regression (MLR), respectively. A total of 160 samples were prepared for the calibration (n=120) and prediction (n=40) sets. Among all the models, the SPA-MLR achieved the best results, where R2p, RMSEP and RPD were 0.950 1, 0.308 7 and 4.476 6 respectively. Results showed that SPA can be effectively used for selecting the effective wavelengths from hyperspectral data. And, SPA-MLR is an optimal modeling method for prediction of apple SSC. Furthermore, less effective wavelengths and simple and easily-interpreted MLR model show that the SPA-MLR model has a great potential for on-line detection of apple SSC and development of a portable instrument.
黄文倩, 李江波, 陈立平, 郭志明. 以高光谱数据有效预测苹果可溶性固形物含量[J]. 光谱学与光谱分析, 2013, 33(10): 2843. HUANG Wen-qian, LI Jiang-bo, CHEN Li-ping, GUO Zhi-ming. Effectively Predicting Soluble Solids Content in Apple Based on Hyperspectral Imaging[J]. Spectroscopy and Spectral Analysis, 2013, 33(10): 2843.