摘要
用遗传算法(GA)和连续投影算法(SPA)分别提取了赣南脐橙叶片高光谱图像的有效信息,对叶绿素的含量用偏最小二乘法(PLS)进行建模定量分析。高光谱图像标定后,提取感兴趣区域(ROI)的平均光谱,用GA和SPA算法分别选出了27和8条特征波长,然后用PLS对叶绿素含量建模。GA-PLS与SPA-PLS模型得到的预测集相关系数分别为0.80和0.83,均方根误差分别为2.45和2.30。结果表明:SPA-PLS模型具有较高的优势,可以结合高光谱技术对赣南脐橙叶绿素含量快速、无损的定量分析。
The hyperspectral effective information of Gannan navel orange was extracted by genetic algorithm （GA） and succes- sive projections algorithm （SPA） for partial least squares （PLS） model. Average spectral was extracted from region of interest （ROI） of hyperspectral images after preprocessing. GA and SPA were used to select 27 and 8 characteristic wavelengths for chlo- rophyll model with PLS. The correlation coefficients and rms error of GA-PLS were 0. 80 and 2. 45, and the correlation coeffi- cients and rms error of SPA-PLS were 0. 83 and 2.30, respectively. Overall results sufficiently demonstrate that SPA-PLS model has a greater advantage which can be combined with hyperspectral technique to be a nondestructive and rapid analytical method.
出处
《光谱学与光谱分析》
SCIE
EI
CAS
CSCD
北大核心
2012年第12期3377-3380,共4页
Spectroscopy and Spectral Analysis
基金
科技部农业科技成果转化资金项目(2011GB2C500008)
国家(863计划)课题项目(2012AA101906)
江西省研究生创新专项资金项目(YC2011-S077)资助
