基于图像处理的玉米品种的种子形态分析及其分类研究

A Study on Basic Morphologic Information and Classification of Maize Cultivars Based on Seed Image Process

【目的】建立中国当代玉米品种的种子重要形态性状的基础信息集。【方法】扫描获得中国当代193个代表性玉米品种各50粒种子的正反面彩色图像,确定大小、形状、纹理和颜色共4大类27个能够系统反映籽粒形态的数量性状,自行编制数字图像处理程序实现了形态特征的自动提取并获得品种各性状的籽粒均值,采用SAS软件的单性状描述统计和单性状或同类性状聚类分析对品种进行评价与分类。【结果】建立了一个含有9650个玉米粒的19300幅彩色图像的数据库,明确了籽粒各个形态性状的变化规律,按变异程度分为大中小3个组别,发现中国当代玉米品种在大多数情况下聚集程度都比较高,筛选出36份(次)特异种质材料。【结论】研究结果为玉米粒的形态研究与应用提供了比较系统全面的数据,数字图像处理作为一种简便快速的作物籽粒形态检测技术将有十分广泛的应用前景。

[ Objective ] The objective of this study is to get systematic and complete information on important seed morphology of maize germplasm resources in China. [ Method ] Digital color images were scanned from 2 side faces each of 50 kernels each of 194 maize cultivars in China. The authors coded computer programs to automatically extract 27 individual morphologic kernel traits belonging to 4 categories of size, shape, texture and color. These quantitative traits are representative of kernel morpha, and means of individual traits were averaged over kernels within cultivars. Measurements were statistically summarized and cluster-analyzed by SAS software to evaluate and classify the cultivars by individual or group morphologic traits. [Result] A database was developed of 19 300 color images of 9 650 kernels. Ranges, means, standard deviations and variation coefficients for each trait were analyzed, and there were large, middle and small variation groups of the morphologic traits. Maize cultivars indicated cluster distributions in most situations. Thirty-six unique germplasm materials were screened out. [ Conclusion] This study provided rich primary data on and insights into morphologic seed aspects in kernel research and application in maize. Digital image process as a technique to measure crop kernel morpha showed promising application perspectives.