无信息变量消除法在筛选南疆红枣总酸近红外特征波长中的应用  

The Application that UVE Method Screens the NIR Characteristic Wa-velengths of Southern Xinjiang Red Jujube Total Cid

作　　者：胡晓男 彭云发 罗华平[1,2,3] 罗雪宁 詹映[1] 代希君 

机构地区：[1]塔里木大学机械电气化工程学院,阿拉尔843300 [2]新疆维吾尔自治区普通高等学校现代农业工程重点实验室,阿拉尔843300 [3]南疆农业机械化研究中心,阿拉尔843300

出　　处：《食品工业》2015年第5期232-235,共4页The Food Industry

基　　金：国家自然基金项目资助(No.10964009和No.11164023)

摘　　要：以南疆红枣总酸的快速无损检测为研究对象,利用SPXY(Sample set partitioning based on joint x-y distances)法来划分校正集样本,应用无信息变量消除法(UVE)对南疆红枣总酸近红外光谱(NIRS)特征变量进行筛选,然后用筛选出的变量建立偏最小二乘(PLS)模型,该模型的预测标准偏差(RMSEP)为0.044 7,预测相关系数为Rp为0.877 8,并将UVE筛选的变量建立的PLS模型与全光谱建立PLS模型结果进行比较。结果表明,SPXY法划分的校正集样本合理;UVE优出选全光谱1 557个变量中的420个变量,建立的PLS模型预测效果要好于全光谱建立的PLS模型,UVE能够有效地选取待测成分的特征波长,建立简化的红枣总酸预测模型,降低模型计算量。Southern Xinjiang red jujube fast nondestructive testing of total acid was taken as the research object. Sample set partitioning based on joint x-y distances （SPXY） was used to find correction collection of samples, and southern Xinjiang jujube total acid near infrared spectrum （NIRS） characteristic variables were screened with uninformative variables elimination （UVE） method. Then, variables were selected to establish a partial least squares （PLS） model. The root mean square error of prediction of the model （RMSEP） was 0.044 7, the predictive correlation coefficient Rp was 0.877 8. PLS model established by UVE screening variables was compared with full spectrum established PLS model results. As the results showed, the corrected set of samples divided with SPXY method was reasonable. With 420 variables that optimized with UVE from the full spectrum of 1 557, the predictive effect of the established PLS model was better than that of the PLS model of full spectrum. UVE effectively selected wavelength characteristics of the component under the test and established a simplified prediction jujube total acid model, reducing the model calculation.

关 键 词：无信息变量消除法 红枣 近红外光谱 特征波长 总酸 

分 类 号：TS255.1[轻工技术与工程—农产品加工及贮藏工程]