高光谱成像技术对鲜枣内外部品质检测的研究

外部缺陷以及内部可溶性固形物的含量对提升鲜枣的采后附加值和鲜枣后续生产加工具有重要的意义，因此，为了实现同时对鲜枣内外部品质进行快速、准确识别，利用高光谱成像技术(450-1,000 nm)对壶瓶枣的“自然损伤”和可溶性固形物含量同时进行检测研究。首先，对光谱数据进行主成分分析(PCA)得到前7个主成分光谱值，对图像数据采用灰度共生矩阵(GLCM)提取到7项图像纹理指标(对比度、相关性、能量、同质性、方差、均值、熵)。然后，分别使用光谱主成分值、图像纹理特征值、以及主成分与纹理特征融合值建立偏最小二乘支持向量机(LS-SVM)模型对壶瓶枣的外部缺陷(“自然损伤”)和内部品质(可溶性固形物含量)进行检测研究。结果表明：使用主成分与纹理特征融合值建立的LS-SVM模型可作为通用模型同时对壶瓶枣内外部品质进行检测研究，其“自然损伤”判别正确率为92.5%，可溶性固形物预测集的预测相关系数(R_p)和预测均方根误差(RMSEP)分别达到了0.944和0.495。表明，采用高光谱成像技术可以建立通用模型同时对壶瓶枣的内外部品质进行检测，该研究为壶瓶枣的无损检测提供了理论参考。

Simultaneous Detection of External and Internal Quality Parameters of Huping Jujube Fruits using Hyperspectral Imaging Technology

Nondestructive detection of external and internal quality parameters of jujube is crucial for improving jujube’s shelf life and industry production. Hyperspectral imaging is an emerging technique that integrates conventional imaging and spectroscopy to acquire both spatial and spectral information from a sample. It takes the advantages of the conventional RGB, near-infrared spectroscopy, and multi-spectral imaging. In this work, hyperspectral imaging technology covered the range of 450～1 000 nm has been evaluated for nondestructive determination of “natural defects” (shrink，crack，insect damage and peck injury) and soluble solids content(SSC) in Huping jujube fruit. 400 RGB images were acquired through four different defect (50 for each stage) and normal(200) classes of the Huping jujube samples. After acquiring hyperspectral images of Huping jujube fruits, the spectral data were extracted from region of interests(ROIs). Using Kennard-Stone algorithm, all kinds of samples were randomly divided into training set (280) and test set (120) according to the proportion of 3∶1. Seven principal components (PCs) were selected based on principal component analysis (PCA), and seven textural feature variables (contrast，correlation，energy，homogeneity，variance，mean and entropy) were extracted by gray level co-occurrence matrix (GLCM). The least squares support vector machine (LS-SVM) models were built based on the PCs spectral, textural, combined PCs and textural features, respectively. The satisfactory results show the correct discrimination rate of 92.5% for the prediction samples, as well as correlation coefficient (R_p) of 0.944 for the prediction set to calculate SSC content based on PCs and textural features. The study demonstrated that hyperspectral image technique can be a reliable tool to simultaneous detection of external (“natural defects”) and internal (SSC) quality parameters of Huping jujube fruits, which provided a theoretical reference for nondestructive detection of jujube fruit.