基于纹理和分形的鲜茶叶图像特征提取在茶树品种识别中的应用  

Research on Feature Extraction of Fresh Tea Image Based on Texture and Fractal and Its Application in Tea Variety Recognition

作　　者：刘自强 周铁军[2] 傅冬和[3] Liu Ziqiang;Zhou Tiejun;Fu Donghe(Hunan Institute Of Traffic Engineering,Hengyang 421001;College of information and intelligence,Hunan Agricultural University,Changsha 410128;College of horticulture,Hunan Agricultural University,Changsha 410128)

机构地区：[1]湖南交通工程学院,湖南衡阳421001 [2]湖南农业大学信息与智能科学技术学院,湖南长沙410128 [3]湖南农业大学园艺学院,湖南长沙410128

出　　处：《中阿科技论坛（中英文）》2021年第6期123-127,共5页China-Arab States Science and Technology Forum

基　　金：湖南省教育厅科学研究项目“鲜茶叶图像特征提取及在茶树品种识别中的应用研究”(17C0589)

摘　　要：本文对鲜茶叶图像格式进行了转换和预处理,并运用灰度共生矩阵和统计矩度量方法提取了鲜茶叶纹理特征,根据多重分形理论知识及相关算法,编写程序用盒子维方法求分形相关特征参数,再分别以六种分类器训练建模,并比较各模型的预测精度。结果表明,SVMKM和随机森林以两类特征建模,运用10折交叉验证、独立预测分类这两种方法的精确度可达89%。In this paper,the image format of fresh tea is converted and pre-processed,and the gray-level co-occurrence matrix and statistical moment measurement method are used to extract the texture features of fresh tea.According to the multifractal theory and related algorithms,a program is written to obtain fractal-related feature parameters using the box dimension method,and then six classifiers are used to train the modeling,and the predicted accuracy of each model is compared.The results show that SVMKM and Random Forest are modeled with two types of features,using 10-fold cross-validation,the accuracy of these two methods of independent prediction and classification can reach 89%.

关 键 词：图像处理 灰度共生矩阵 多重分形 特征筛选 分类识别 

分 类 号：TP391.41[自动化与计算机技术—计算机应用技术]