基于小麦籽粒高光谱特征的品种鉴别研究

RESEARCH ON WHEAT GRAIN VARIETIES IDENTIFICATION BASED ON HYPERSPECTRAL CHARACTERISTICS

为探究高光谱成像技术在小麦籽粒品种鉴别等的研究,以15种品种已知的小麦籽粒为对象,分别采集样品的光谱和图像信息。运用主成分分析法优选三个特征波长(447nm、615nm、955nm),提取特征波长下小麦籽粒图像的形态特征(面积、周长、圆度、长轴长度、短轴长度)和纹理特征(均值、标准差、熵)。应用Bayes判别分析法进行多元统计分析,建立判别函数判别回代准确率为99.9%,交叉验证的准确率为98%,模型的判别效果良好。研究表明利用高光谱成像技术结合Bayes判别分析的方法可用于小麦籽粒品种的鉴别。

This study aims at exploring how to identify wheat grain varieties by hyperspectral image technology.The samples were fifteen known wheat grain varieties, which were collected for hyperspectral image and spectral data respectively. Using principal component analysis（PCA） method to find out three characteristic wavelengths（447nm, 615nm, 955nm）, after that extract the morphological characteristics（area, length, roundness, major axis length, minor axis length） and texture characteristics（mean, smoothness, entropy） of wheat grain from hyperspectral image based on characteristic .wovelengths. The Bayes discriminant anal-ysis method is used to do multivariate statistical analysis of discriminative accuracy of back substitution is 99.9% good. The results show that the method which combines and build the discrimination function. The result and cross validation accuracy is 98%, which is hyperspectral image technology with Bayes dis-criminant analysis can be used to identify wheat varieties.