苹果糖度近红外光谱小波去噪和iPLS建模  

Near Infrared Determination of Sugar Content in Apples Based on Multi- resolution Decomposition and Interval Partial Least Square (iPLS) Method

作　　者：邹小波[1] 赵杰文[1] 夏蓉[2] 孙乐六[2] 

机构地区：[1]江苏大学农产品加工研究所 [2]江苏恒顺集团有限公司

出　　处：《农业机械学报》2006年第6期79-82,共4页Transactions of the Chinese Society for Agricultural Machinery

基　　金：国家自然科学基金资助项目(项目编号:30370813);高等学校博士学科点专项科研基金资助项目(项目编号:20040299009)

摘　　要：为了提高苹果近红外光谱糖度预测模型的精度,利用多尺度小波去噪法对苹果近红外光谱进行了预处理,并用改进后的间隔偏最小二乘法(iPLS)建立预测模型。应用结果表明,多尺度小波去噪法滤除了原始光谱中的部分噪声,但又保留了原光谱中的主要信息。运用间隔偏最小二乘法对预处理后的光谱建模,其校正时的相关系数rc和校正均方根误差RMSEC分别为0.9635和0.3026,预测时的相关系数rp和预测均方根误差RMSEP分别为0.9214和0.4113,主因子数为5个。结果表明,用多尺度小波去噪和间隔偏最小二乘法所建立的苹果糖度模型不但精度有所提高,而且更加简洁、数据运算量也更少。To improve the prediction model of sugar content, multi-resolution decomposition was used to preprocess the near infrared （NIR） spectra of apples. Compared with those original spectra, after pretreatment apple spectra were smoother, but their shape showed no much difference. This indicated that the major information in apple spectra could be reserved while noise was removed by multi-resolution decomposition method. An evolution of interval partial least square （iPLS） method was proposed and used to establish the calibration models of sugar content against apple spectra after multi-resolution decomposition pretreatment. The optimum iPLS calibration model was obtained with 5 factors, the correlation coefficient （rp） of 0. 963 5 with the root mean square error of calibration （RMSEC） of 0. 302 6 and the prediction coefficient （re） of 0. 921 4 with the root mean square error of prediction （RMSEP） of 0. 411 3. Compared with whole spectra data model, the iPLS model could not only improve precision, but also simplify the model.

关 键 词：近红外光谱 多尺度小波 间隔偏最小二乘法 噪声 苹果 糖度 

分 类 号：O657.33[理学—分析化学] S123[理学—化学]