Abstract: Incompletely closed glumes, germination on panicle and disease were three characteristics of hybrid rice seed causing poor seed quality. The typical images of hybrid rice seeds with germination, severe disease, incompletely closed glumes, spot disease, fine fissure and normal seeds were made with a machine vision system. Each image could be processed to get the feature values of seed such as length, width, ratio of length and width, area, solidity and hue. Then all images of normal seeds were calculated to draw the feature value ranges of each variety. Finally, an image information base which stores typical images and feature values of each category was established.

Key words: hybrid rice seed, image processing system, image information base, characteristics, quality inspection, methodology

摘要： 为提供机器视觉检验杂交稻种时的检验策略及其各类基本图像信息，将杂交稻种特征特性分为芽谷、重霉、裂颖、点霉、微裂和正常等六类进行检验处理，利用自行研制的数字图像系统采集金优402、汕优10号、中优207、嘉优99、Ⅱ优3207共5个杂交稻种的样品图像，在此基础上运用开发的相关算法软件对每幅图像进行计算，从而获得各类稻种的长、宽、长宽比、面积、紧密度、色调等外观特征参数值范围，进而建立了5个杂交稻种的图像信息库。图像信息库可显示各杂交稻种的典型图像和相关特征值信息，有助于明确其特征特性的分类界限，为检验提供阈值选择的参考，并支持大量图片的管理、新品种的添加等，是机器视觉检验稻种的基础。

关键词: 杂交稻种, 图像信息库, 图像采集系统, 特征特性, 质量检验, 方法