基于高光谱成像技术的薯麦混合粉淀粉含量检测  

Detection of Starch Content in Potato-Wheat Flour Mixtures Based on Hyperspectral Imaging

作　　者：岳明慧 张姗姗 张静 印祥 李宏军[1] 马成业[1] Yue Minghui;Zhang Shanshan;Zhang Jing;Yin Xiang;Li Hongjun;Ma Chengye(School of Agricultural Engineering and Food Science,Shandong University of Technology,Zibo 255000)

机构地区：[1]山东理工大学农业工程与食品科学学院,淄博255000

出　　处：《中国粮油学报》2023年第12期197-202,共6页Journal of the Chinese Cereals and Oils Association

基　　金：山东省重点研发计划(2022CXGC010604),山东省重点研发计划项目(2019JZZY010734)。

摘　　要：为快速无损检测薯麦混合粉中的淀粉含量,在近红外(900~2500 nm)的光谱范围采集样品的高光谱图像,通过对比原始光谱和七种不同预处理后的光谱数据建立偏最小二乘回归(PLSR),支持向量机回归(SVMR)模型效果,确定多元散射校正法(MSC)为最佳预处理方法。采用竞争性自适应加权法(CARS)和连续投影算法(SPA)提取特征波长,建立PLSR和多元线性回归(MLR)模型,结果表明SPA显著的降低了模型的复杂度,SPA-MLR模型检测集的决定系数(R_(P)^(2))为0.9181,检测集的均方根误差为0.4214%。To rapidly and non-destructively detect the content of starch in potato-wheat flour mixtures,hyperspectral images of the samples were collected in the near infrared(900~2500 nm)spectral range,and partial least square regression(PLSR)was established by comparing the original spectrum with seven different pretreated spectral data.Support vector machine regression(SVMR)model effect was supported,and multiple scattering correction(MSC)was determined as the best preprocessing method.Competitive adaptive weighted sampling(CARS)and successive projections algorithm(SPA)were used to extract the characteristic wavelength.PLSR and multiple linear regression(MLR)models were established.The results indicated that SPA significantly reduced the complexity of the model,and the determination coefficient of prediction(R_(P)^(2))of the SPA-MLR model detection set was 0.9181 with the root mean square error of prediction(RMSEP)of 0.4214%.

关 键 词：薯麦混合粉 淀粉含量 高光谱成像技术 无损检测 可视化 

分 类 号：TS213.2[轻工技术与工程—粮食、油脂及植物蛋白工程] O657.3[轻工技术与工程—食品科学与工程]