基于高光谱成像技术的包衣甘蓝种子色度检测  

Detection of Coated Cabbage Seeds Color Based on Hyperspectral Imaging Technology

作　　者：吴龙国[1,2] 马玲 张瑶[1] 田雨[1] 朱彦哲 张祎洋 WU Long-guo;MA Ling;ZHANG Yao;TIAN Yu;ZHU Yan-zhe;ZHANG Yi-yang(College of Enology and Horticulture,Ningxia University,Yinchuan 750021,China;Ningxia Modern Protected Horticulture Engineering Technology Research Center,Yinchuan 750021,China)

机构地区：[1]宁夏大学葡萄酒与园艺学院,宁夏银川750021 [2]宁夏现代设施园艺工程技术研究中心,宁夏银川750021

出　　处：《分析测试学报》2025年第3期454-463,共10页Journal of Instrumental Analysis

基　　金：国家重点研发计划子课题专项(2021YFD1600302-3)。

摘　　要：采用高光谱成像技术对包衣过程中4个不同品种的包衣甘蓝种子,3个不同浓度包衣剂处理的包衣甘蓝种子的包衣颜色均匀性以及包衣颜色深浅进行分析。提取出240个种子样本的平均光谱反射率,通过4种预处理方法对原始光谱进行预处理和优化,然后用竞争自适应重加权算法(CARS)、连续投影算法(SPA)、无信息变量消除变换法(UVE)、遗传偏最小二乘法(GAPLS)4种方法提取特征波长。基于优选的特征波长建立了偏最小二乘回归(PLSR)、多元线性回归(MLR)以及主成分回归(PCR)模型。结果表明:羽衣甘蓝种子的包衣效果最明显,佳香口感型甘蓝次之,中甘15和紫甘蓝的包衣效果接近;优选Baseline法对色度值L*进行预处理,Normalize法对色度值a进行预处理,SNV法对色度值b进行预处理;GAPLS法提取的特征波长用于建立L^(*)、b的定量预测模型,UVE法提取的特征波长用于建立色度a值的定量预测模型。PLSR建立的L*模型效果最优(R_(C)=0.814,Rp=0.640;RMSEC=1.150,RMSEP=1.852);MLR建立的色度a值模型效果更优(R_(c)=0.981,Rp=0.964;RMSEC=2.563,RMSEP=3.243);PCR建立的色度b值模型效果最优(R_(C)=0.917,Rp=0.913;RMSEC=2.552,RMSEP=2.589)。研究结果可为种子色度的在线监测提供技术支撑。Hyperspectral imaging technology was used to analyze the coating color uniformity and the depth of coating color of coated cabbage seeds of 4 different varieties and coated cabbage seeds treated with coating agents at 3 different concentrations during the coating process.The average spectral reflectance of 240 seed samples was extracted.The original spectra were preprocessed and optimized through 4 preprocessing methods.Then,4 methods including the competitive adaptive reweighted sampling algorithm(CARS),successive projections algorithm(SPA),uninformative variable elimination transformation method(UVE),and genetic algorithm partial least squares algorithm(GAPLS)were used to extract the characteristic wavelengths.Partial least squares regression(PLSR),multiple linear regression(MLR)and principal component regression(PCR)were established based on the optimized characteristic wavelengths.The results showed that the coating effect on kale seeds was the most obvious,followed by that on Jiaxiang taste-type cabbage,and the coating effects on Zhonggan 15 and purple cabbage were close.The Baseline method was preferably used to preprocess the chromaticity value L^(*),the Normalize method was used to preprocess the chromaticity value a,and the SNV method was used to preprocess the chromaticity value b.The characteristic wavelengths extracted by the GAPLS method were used to establish the quantitative prediction models for L*and b,the characteristic wavelengths extracted by the UVE method were used to establish the quantitative prediction model for the chromaticity value a.The L*model established by PLSR had the best effect(Rc=0.814,Rp=0.640;RMSEC=1.150,RMSEP=1.852),the model for the chromaticity value a established by MLR had a better effect(Rc=0.981,Rp=0.964;RMSEC=2.563,RMSEP=3.243),and the model for the chromaticity value b established by PCR had the best effect(Rc=0.917,Rp=0.913;RMSEC=2.552,RMSEP=2.589).The research can provide technical support for the online monitoring of seed chromaticity.

关 键 词：高光谱成像 包衣甘蓝种子 色度 检测 

分 类 号：O657.3[理学—分析化学] S482.2[理学—化学]