不知火杂柑可溶性固形物在线检测模型建立及优化

发布时间：2018-06-16 23:48

本文选题：不知火杂柑 + 近红外漫透射光谱　；参考：《光谱学与光谱分析》2017年05期

【摘要】：应用近红外漫透射光谱检测技术对不知火杂柑的可溶性固形物(SSC)进行在线检测具有十分重要的意义。研究变量筛选方法对不知火杂柑可溶性固形物在线检测模型的影响,为实现其快速、准确的在线检测分级奠定基础。实验把形状不整、内藏瓤瓣的不知火杂柑作为研究对象,选取560~930nm的光谱,采用偏最小二乘法(PLS)建立不知火杂柑可溶性固形物的在线检测模型,并讨论不同的光谱预处理方法(卷积平滑(S-G)、一阶微分(1st derivatives)等),不同的变量筛选方法(移动窗口偏最小二乘法MWPLS、遗传算法GA、连续投影SPA)对PLS所建预测模型性能的影响。经对比,多元散射校正(MSC)能有效地消除光散射的影响,遗传算法能大大地降低了建模的波长点数,缩短了建模时间,改善模型预测精度。其最优PLS模型的RP=0.956,RMSEP=0.380,RC=0.967,RMSEC=0.340。实验表明在线检测不知火杂柑的可溶性固形物是完全可行的。
[Abstract]:It is very important to use near infrared diffuse transmission spectroscopy (NIR) to detect the soluble solids (SSCs) of mandarin. The effect of variable selection method on the on-line detection model of soluble solids in unintelligible mandarin was studied, which laid a foundation for fast and accurate on-line detection and grading. In this experiment, the incompletely shaped and flaps of unintelligible mandarin were chosen as the research object, the spectrum of 560~930nm was selected, and the on-line detection model of soluble solids was established by partial least square method (PLS). The effects of different spectral pretreatment methods (convolution smoothing S-Gn, first order differential derivation) and different variable screening methods (moving window partial least square method MWPLS, genetic algorithm GA, continuous projection spa) on the performance of PLS prediction model are discussed. By comparison, multivariate scattering correction (MSCM) can effectively eliminate the influence of light scattering. Genetic algorithm can greatly reduce the number of wavelength points, shorten the modeling time and improve the prediction accuracy of the model. The optimal PLS model of RP0. 956 RMS EPN 0.380 RCU 0.967N RMSEC 0.340. The experimental results show that it is feasible to detect soluble solids of Citrus mandarinus on-line.
【作者单位】：华东交通大学机电学院光机电技术及应用研究所;
【基金】：国家“863”高技术研究发展计划项目(2012AA101906) 赣鄱英才555工程领军人才培养计划项目(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) 江西省优势科技创新团队建设计划项目(20153BCB24002)资助
【分类号】：O657.33;TS255.7

【相似文献】