基于高光谱技术的长枣内外品质同时检测

Simultaneous detection of external and internal quality parameters of long jujubes using hyperspectral imaging technology

摘要:

采用近红外高光谱成像技术,对长枣表面轻微损伤 和果肉硬度进行无损检测。在970～ 1630nm波 长范围内对高光谱图像数据进行主成分分析(PCA),得到第3主成分图像最适合检测长枣表面 损伤。波段比(BR)算法中, 选取1387nm和1 229nm两个波段的图像进行比值运算, 采用 1455nm单波 段图像构建掩膜作用于比值图像, 最后对图像进行阈值分割和形态学变换完成损伤区域的特征提取。BR算法检测长枣轻微损 伤的准确率达 到91.5%。对反射光谱进行多元散射校正(MSC)后与长枣果肉硬度值进 行回归分析,选择相关系数较大的5个特 征波长作为BP神经网络输 入,建立果肉硬度预测模型。预测集相关系数R和均 方根误差(RMSEP)分别为0.904和15.163。研究结果表明,利用高光谱成像技术可以实现长枣内外品质同时检 测。

Abstract:

Near-infrared hyperspectral imaging technology is adopted to take the n ondestructive detection for the subtle bruise and flesh firmness of the long jujubes.Hyperspectral imaging data in the range of 970－1630nm is analyzed by the principal component analysis (PCA),and the third principal compo nent image (PC3) is the most suitable for identifying the bruise of long jujubes .It has a ratio operation fo r the image by choosing the bands 1387nm and 1229nm with the band ratio algorithm,and adopts a singl e ba nd image of 1455nm to build the mask which acts on the ratio image.Lastly,the characteristic extraction about the brui se area is finished by threshold dividing and morphological transforming.This algorithm shows that the subtle bruise could be detected with an accuracy of 91.5%.It has a regression analysis between the reflection spectrum after using multiple scatter correction and the flesh firmness of jujubes,chooses five characteristic wavelengths with bigger correlation coefficient as inputs of the BP neural network and builds a prediction model based on the flesh firmness .The correlation coefficient and the root mean square error of the prediction model are 0.9042and 15. 1631,respectively.This research demonstrates that it is feasible to detect external and internal quality parameters of long jujubes simu ltaneously with hyperspectral imaging technology.