摘要
为了提高近红外光谱技术在梨的可溶性固形物含量(SSC)检测中的精度和稳定性,对采集的原始光谱进行标准归一化(SNV)预处理,采用联合区间偏最小二乘法(SiPLS)建立了SSC的预测模型;通过交互验证法确定了模型的主成分因子数,以预测时的相关系数(Rp)和预测均方根误差(RMSEP)作为评价指标对模型预测结果进行了分析,并与经典偏最小二乘(PLS)模型、间隔偏最小二乘(iPLS)模型进行了比较。结果表明,利用SiPLS所建的预测模型的最优组合包含21个光谱区间并联合4个子区间和15个主成分因子,其预测集的相关系数和预测均方根误差分别为0.9633和0.203;说明利用近红外光谱结合SiPLS算法可以准确、无损检测梨中可溶性固形物含量。
In order to improve the precision and stability of determination of soluble solids content（SSC） in pear by FT-NIR spectroscopy,the collected original spectra of was pretreated by standard normalization（SNV）,and prediction medel of SSC was established by synergy interval partial least-squares（SiPLS）.The number of SiPLS components was confirmed by the cross-validation,the predict results of SiPLS model were analyzed with correlation coefficient（Rp） and the root mean square error of prediction（RMSEP） as evaluation index,compared with the classical PLS model and interval PLS（iPLS） model.Results showed that the optimal prediction model by SiPLS contained 21spectral interval combined with 4 subinterval and principal component factor was 15,and Rp and RMSEP of prediction were 0.9633 and 0.203,respectively.It is concluded that NIR spectroscopy combining with SiPLS can be applied to accurate and lossless determination of the SSC in pear.
出处
《光谱实验室》
CAS
2013年第1期68-72,共5页
Chinese Journal of Spectroscopy Laboratory
基金
国家自然科学基金资助项目(50775151)
