基于叶片形态特征的葡萄品种自动识别

Automatic Identification of Grape Varieties Based on Leaf Morphological Characteristics

葡萄在全世界种植范围较广,葡萄品种识别对葡萄资源统计、新品种鉴定及遗传资源保护都有重要意义。以成熟叶片为基本材料,不变矩、面积、周长、主叶脉长度和叶柄长度等11维特征向量为基础数据,采用叶片特征选择、图像处理、特征值提取、建立分类模型等方法,运用国际葡萄与葡萄酒组织形态分类方法和改进的欧式距离等技术开发一个基于葡萄叶片数字图像的葡萄品种自动识别软件。共测试17种酿酒和3种野生葡萄,识别率达87%。该方法具有自动化、识别速度快、花费低、省时省力等特点,可应用于葡萄病虫害的识别、杂交新品种双亲的判定等。

Grapes have been widely cultivated in the world.Identification of grape varieties is meaningful for grape resources statistics,the identification of new varieties and protection of grape genetic resources.In this paper,we presented our grape classifier based on 11-dimension feature vectors including Hu＇s moment invariants,perimeter,area,the length of the main vein and petiole of grape mature leafs,through the following steps： determination of leaf features,image processing,calculating feature vectors from leaf images,and establishing classification model.Our aim was to develop an automatic identification software based on grape leaf＇s digital image,combining OIV morphological taxonomy with Euclidean distance enhanced by probability.Our software achieves a correct recognition rate higher than 87% for seventeen kinds of vine grapes and three kinds of wild grapes.The recognition was entirely performed by the software.It is automatic,fast and cheap without any specific device.This technique can be applied in identification of grape diseases and insect pests as well as the parent determination of new hybrid grapes varieties.