基于WLLE和SVM的植物叶片图像识别方法  

Recognition method of plant leaves based on WLLE and SVM

作　　者：丁娇[1] 梁栋[1] 阎庆[1] 

机构地区：[1]安徽大学电子信息工程学院,安徽合肥230039

出　　处：《安徽大学学报（自然科学版）》2013年第4期61-67,共7页Journal of Anhui University(Natural Science Edition)

基　　金：国家自然科学基金资助项目(61172127);高等学校博士学科点专项科研基金资助项目(20113401110006);安徽省自然科学基金资助项目(1208085QF104);安徽省高校优秀青年人才基金资助项目(2012SQRL017ZD)

摘　　要：针对局部线性嵌入(LLE)算法易受噪声影响,以及最近邻分类器不能有效识别植物叶片图像,提出一种基于加权局部线性嵌入(WLLE)和支持向量机(SVM)的植物叶片图像识别方法.首先利用WLLE算法对预处理后的含有高斯噪声叶片图像进行特征提取,然后采用SVM分类机制对叶片图像进行训练和识别,最后在真实的植物叶片图像数据库中提取植物叶片图像进行分类实验.实验结果表明该方法能提高叶片图像的分类率.In general, noise could influence the algorithm of LLE, and nearest neighbor classifier couldn＇ t recognize plant leaf images effectively, a recognition method of plant leaves based on weighted locally linear embedding and support vector machine was proposed. The features of the preprocessing plant leaf images with noise were extracted by use of WLLE, and leaf feature sets were trained and recognized by the classification methods of SVM, at last, plant leaf images from the real plant leaf image database were used to take classification experiment. Experimental results showed that the proposed method improved the classification accuracy of plant leaf images.

关 键 词：流形学习 局部线性嵌入 加权局部线性嵌入 特征提取 支持向量机 植物叶片识别 

分 类 号：TP391.41[自动化与计算机技术—计算机应用技术]