基于叶片数字纹理特征自动识别胡颓子属植物  

Automatic Identification of Elaeagnus L. Based on Leaf Digital Texture Feature

作　　者：王雷宏[1] 陈永生[1] 郑玉红[2] Wang Leihong;Cheng Yongsheng;Zheng Yuhong(School of Forestry and Landscape of Architecture,Anhui Agricultural University,Hefei 230036;Institute of Botany,Jiangsu Province and the Chinese Academy of Sciences,Nanjing 210014)

机构地区：[1]安徽农业大学林学与园林学院,合肥230036 [2]江苏省中国科学院植物研究所,南京210014

出　　处：《中国农学通报》2020年第11期20-25,共6页Chinese Agricultural Science Bulletin

基　　金：国家自然科学基金资助项目“不同林龄序列亚热带常绿阔叶林地下碳氮耦合循环特点”(31370626)。

摘　　要：为了探索胡颓子属植物叶片的数字纹理特征变异规律,对苏、浙、皖地区常见的8种胡颓子属植物,提取了基于灰度共生矩阵的叶片纹理参数,分析了叶片纹理参数的种内、种间变异规律,并构建KNN分类模型。结果表明:同种不同地理来源的标本间全部纹理参数是极显著差异,不同种之间仅某一纹理参数有显著差异;随机取132个样本作为训练集,35个作为测试集,构建KNN分类模型,K=6时,正确识别率达到了93.75%。对于特定分布区内的几个胡颓子属植物,叶片数字纹理具有分类识别意义,可用于构建分类模型。This study aims to explore the variation pattern of leaf digital texture feature of Elaeagnus L.. The leaf texture parameters were extracted based on Gray-level co-occurrence matrix from the eight species of Elaeagnus L. from Zhejiang, Jiangsu, and Anhui Province. The variation pattern of leaf texture parameters was analyzed within and among species. KNN classification model was established. The results showed that all texture parameters of the same species from different geographical sources had highly significant differences,while only one texture parameter had significant difference between different species. The KNN classification recognition model was constructed by 132 random samples as train data, 35 random samples as test data. The correct recognition rate of this model was 93.75% at K=6. The leaf digital texture has the significance of classification recognition to some species of Elaeagnus L. in a certain distribution range, and it can be used to construct k-nearest neighbor classification model.

关 键 词：植物识别 胡颓子属 叶片 纹理参数 最近邻分类器 

分 类 号：S718.49[农业科学—林学]