摘要
近红外光谱分析技术是近年迅速发展的一门绿色分析技术,具有快速、准确、无损伤检测的特点,正越来越广泛地应用于水果内部品质的无损检测。为此,应用近红外漫反射光谱技术无损检测芒果糖度,光谱数据经一阶微分和Savitzk-Golay预处理后,分别采用主成分回归法(PCR)和偏最小二乘法(PLS)建立芒果糖度近红外分析模型,当主成分数为10时,预测相关系数R分别为0.838 69,0.976 59,RMSEC分别为1.628 4,0.205 8,MEP为1.235 0,0.738 3;采用BP神经网络对70个样品光谱进行训练,建模集相关系数达到0.983 3,而对36个预测样品集的相关系数只有0.663 9;调整建模集和预测集样品,相关系数增大到0.683 6,平均相对误差由10.336 9%降到8.057 6%。研究结果表明,利用近红外漫反射光谱技术对芒果糖度进行无损检测是可行的。
Near infrared spectroscopy nondestructive detecting technique is a green analytical technique,which has developed faster in recent years,has fast,accurate and non-destructive merits.It can evaluate internal quality of fruit simultaneously without destroying samples.After first derivative D1log（1/R） and Savitzky-Golay smoothing were used to preprocess the primitive spectrum of mango respectively,the analysis model of NIR mango suger content by using PCR and PLS methods was established.When the principal component count is 10,two models＇ sugar predictive power achieve the best predict results for：R is 0.838 69 and 0.976 59,RMSECV is 1.628 4 and 0.205 8,MEP is 1.235 0 and 0.738 3.Using BP neural networks trains 70 mango samples,the results show that the correlation coefficients（R） of calibration model is 0.983 3.And using this model predict the 36 predicted samples,the results is R=0.663 9,and then adjust the calibration samples and predicted samples,R increases to 0.683 6,the average relative error reduces from 10.336 9% to 8.057 6%.The study shows that the near infrared diffuse reflectance technique was feasible to nondestructively measure the sugar content of mango.
出处
《农机化研究》
北大核心
2013年第1期177-180,共4页
Journal of Agricultural Mechanization Research
基金
广东省科技计划项目(00120021110305004)
关键词
近红外漫反射
芒果
糖度
无损检测
near-infrared diffuse reflectance
mango
sugar content
non-destructive detection
