中文摘要：

为了简化黄瓜叶片叶绿素光谱模型和提高模型预测精度,采用联合区间偏最小二乘法（SiPLS）结合净分析物法（NAS）提取近红外光谱的特征信息,建立了黄瓜叶片叶绿素光谱模型。收集了110片新鲜黄瓜叶片,用近红外光谱仪采集光谱数据后立刻用化学分析方法测定叶绿素含量。原始光谱经过SNV预处理和子区间总数优化后,将全光谱均匀划分为29个子区间,用联合区间偏最小二乘法优选出4个特征子区间,在上述特征子区间的基础上,用净分析物法分离光谱中同叶绿素相关的光谱信息,并结合线性回归法建立了叶绿素光谱模型。模型对应的校正集相关系数Rc、校正均方根误差、预测集相关系数Rp和预测均方根误差分别为0.947 2、0.079 5 mg/g、0.925 0和0.090 6 mg/g。结果表明：联合区间偏最小二乘法结合净分析物法能够有效提取叶绿素的特征光谱信息,提高模型精度的同时降低其复杂度。

英文摘要：

To improve and simplify the prediction model of chlorophylls content of cucumber leaves,synergy interval partial least squares（SiPLS） and net analyte signal（NAS） were combined to search for optimized informative spectral wavelengths about chlorophylls content from NIR spectra of cucumber leaves,then spectral model was developed on the basis of chlorophylls contents.One hundred and ten cucumber leaves were selected to collect NIR spectra and chlorophylls content according to chemical analysis.The spectra were preprocessed by SNV method and divided into 29 intervals,among which 4 subsets,i.e.No 3,4,5,15 were selected by SiPLS.Then NAS was used to characterize the net signals of chlorophyll from cucumber leaf spectra which were used for regression variables of NAS model.The NAS calibration model was obtained with the correlation coefficient Rc of 0.947 2,root mean square error of calibration of 0.079 5 mg/g,the prediction coefficient Rp of 0.925 0 and root mean square error of prediction of 0.090 6 mg/g.It proves that SiPLS-NAS could determine optimal variables in NIR spectra and improve the accuracy of model.