基于高光谱成像技术检测干制红枣VC和总糖含量

Detection of V_(C) and Total Sugar Contents of Dried Jujube Based on Hyperspectral Imaging Technology

为实现对不同干燥时间红枣VC和总糖含量的无损检测,提出了基于高光谱成像技术的无损检测方法。在本研究中,采集不同干燥时间红枣的光谱数据,依次通过光谱预处理、特征波长筛选(连续投影算法)后获取特征波长,通过灰度共生矩阵提取特征波长下的灰度图像纹理数据;分别建立基于特征波长、图像特征以及两者融合特征的偏最小二乘(partial least squares,PLS)预测模型。结果表明,VC和总糖含量预测模型中,分别优选出无预处理和标准化(Autoscale)预处理为最佳方法;最优预测模型分别为基于特征波长建立的和基于数据融合建立的PLS预测模型,对应的预测集决定系数R^(2)_(P)分别为0.930和0.883,预测集均方根误差(root mean square error of prediction,RMSEP)分别为30.439 mg/100 g和0.0400 g/g,模型剩余预测残差均大于2.5,具有较好的预测效果。研究结果表明高光谱成像技术可以用于红枣干燥过程中VC和总糖含量的无损检测。

In order to realize the non-destructive detection of VC and total sugar contents of jujube at different drying time,a non-destructive detection method based on hyperspectral imaging technology was proposed.In this study,the spectral data collected from jujube with different drying time were processed with spectral pretreatment and feature wavelengths screening(successive projections algorithm)to obtain the feature wavelengths,the texture data of the grayscale images at the feature wavelengths were extracted by gray level co-occurrence matrix(GLCM).The partial least squares(PLS)model was built based on feature wavelengths,image feature,and fusion features.As for the prediction models of VC content and total sugar content,the results showed that no pre-processing and Autoscale pre-processing were the best methods of preprocessing,and the PLS prediction models based on feature wavelengths and data fusion had the best prediction performance with coefficient of determination for prediction R^(2)_(P)of 0.930 and 0.883,and the mean square error for prediction(RMSEP)of 30.439 mg/100 g and 0.0400 g/g,respectively.The residual predictive deviation of the model was over 2.5,indicating a good prediction performance.The results showed that hyperspectral imaging technology could be used for non-destructive detection of VC and total sugar contents during the drying process of jujube.