摘要
以晚熟脐橙为试材,采用近红外光谱技术与常规检测分析相结合的方法,对比和评价了基于果面和果汁光谱信息的脐橙可溶性固形物(TSS)含量预测模型精度,并筛选了可溶性固形物预测特征光谱.通过对果面和果汁原始光谱的多元散射校正(MSC)预处理,利用偏最小二乘法(PLS)分别建立了TSS预测模型,其中,当果面光谱主因子为5时,其对于可溶性固形物预测相关系数为最大(R=0.836 7)、预测均方根误差(RMSEP)为最小(RMSEP=0.490 3);而当果汁光谱主因子为8时,其对果汁可溶性固形物的预测相关系数为最大(R=0.905 8)、预测均方根误差为最小(RMSEP=0.523 6).采用联合区间偏最小二乘法(siPLS)对果面和果汁光谱特征波段组合进行筛选,获得果面光谱建模特征波段组合为1 000~1 107,1 750~1 857,2 071~2 177和2 178~2 284nm,建立的校正集和预测集模型相关系数分别为0.946 2和0.902 0,RMSECV为0.359 6,RMSEP为0.430 9;获得用于果汁光谱建模的特征波段组合为1 000~1 125,1 251~1 375,1 376~1 500和1 626~1 750nm,校正和预测模型相关系数分别为0.989 4和0.959 6,RMSECV为0.163 1,RMSEP为0.312 8.结果表明:试验所筛选出的果面和果汁近红外光谱特征波段组合建立的校正模型,均可用于晚熟脐橙TSS含量的无损检测,果汁光谱对于甜橙果实固形物含量预测精度高于果面光谱,近红外光谱技术用于橙汁固形物检测是可行的.
The total soluble solid(TSS)prediction accuracy for late-mature navel orange using near infrared spectroscopy(NIRS)was evaluated and compared between the spectra collected from fruit surface and juice.The characteristic bands from fruit surface and juice for TSS prediction were screened.The raw spectra of fruit surface and juice were preprocessed using multiplicative scatter correction(MSC),and the TSS prediction models were established using partial least square(PLS),respectively.When the numbers of principal components reached 5and 8,the predicted correlation coefficient(R=0.836 7and 0.905 8) and root mean square error of prediction(RMSEP=0.490 3and 0.523 6)were optimal for fruit surface and juice,respectively.Characteristic bands of fruit surface and juice spectra were selected using synergy interval partial least square(siPLS).The characteristic band combinations including 1 000~1 107nm, 1 750~1 857nm,2 071~2 177nm and 2 178~2 284nm were optimal for the TSS modeling of fruit surface,the correlation coefficient of calibration set and prediction set were 0.946 2and 0.902 0,respectively,and RMSECV and RMSEP were 0.359 6was 0.430 9,respectively.The characteristic band combinations including 1 000~1 125nm,1 251~1 375nm,1 376~1 500nm and 1 626~1 750nm were optimal for the TSS modeling of juice,the correlation coefficient of calibration set and prediction set were 0.989 4 and 0.959 6,respectively,and RMSECV and RMSEP were 0.163 1and 0.312 8,respectively.The above results showed that the models established with characteristic band combinations could be used to non-destructively examine TSS and real-time monitor the quality of juice for late-maturing navel orange.
出处
《西南大学学报(自然科学版)》
CAS
CSCD
北大核心
2014年第3期122-129,共8页
Journal of Southwest University(Natural Science Edition)
基金
国家高技术发展计划863课题(2012AA101904)
国家国际科技合作专项项目(2013DFA11470)
重庆市国际科技合作项目(CSTC2011gjhz80001)
国家科技支撑计划课题(2012BAD35B08)
国家星火计划项目“重庆现代柑桔产业技术集成与产业化”(2012GA811001)
重庆市科技攻关项目(cstc2011AC1021)
关键词
鲍威尔脐橙
近红外漫反射光谱
可溶性固形物
联合区间偏最小二乘法
Powell navel orange
near infrared spectrum
total soluble solid(TTS)
synergy interval partial least square(siPLS)