基于SVM的小麦叶部病害识别方法研究  

Identification of Wheat Leaf Diseases Based on SVM Method

作　　者：余秀丽[1] 徐超[1] 王丹丹[1] 张卫园 屈卫锋 宋怀波[1] 

机构地区：[1]西北农林科技大学机械与电子工程学院,陕西杨凌712100

出　　处：《农机化研究》2014年第11期151-155,159,共6页Journal of Agricultural Mechanization Research

基　　金：国家"863"计划项目(2013AA10230402);国家自然科学基金项目(31000670);中央高校基本科研业务经费资助项目(QN2011031)

摘　　要：为了准确识别小麦叶部常见病害,为小麦病情诊断和发展状况判断提供科学依据,设计并实现了一种基于SVM(Support Vector Machine)的小麦叶部常见病害识别方法。该方法可以实现对小麦白粉病、条锈病和叶锈病的准确识别。首先,基于中值滤波法和K均值聚类算法,实现了图像的去噪及病斑分割;然后,提取了病斑区域形状特征和纹理特征;最后,利用SVM算法对小麦叶部病害进行了分类识别。随机试验结果表明,利用所提取的特征可以有效地实现小麦叶部常见病害的识别,基于形状特征的综合识别率可达99.33%;利用SVM算法进行小麦病害叶片识别是有效的、可行的。该方法对于农作物病害智能识别的推广具有较好的借鉴意义。In order to identify wheat blade diseases accurately and provide scientific basis for wheat diseases diagnosis,an identification method of wheat blade disease based on SVM was presented.Three wheat blade diseases,such as Powdery mildew,Strip rust disease and Blade rust disease,were recognized accurately via this method.First,image de-noising and spot segmentation were achieved by using median filter method and K-means clustering algorithm.Then,two kinds of disease spot features were extracted,including shape features and texture features.Finally,disease blades were classified and identified using SVM algorithm.Experiment results showed that the presented method which took advantage of the disease spot features could identify wheat blade diseases efficiently,and the comprehensive identification rate of shape features based method reached 99.33%,which showed that it was very effective and practicable for identifying wheat blade diseases by using SVM algorithm.The presented method also provides reference for intelligent identification of crop diseases in the future research.

关 键 词：小麦叶片 病斑识别 特征提取 支持向量机 

分 类 号：TP391.4[自动化与计算机技术—计算机应用技术]