基于SVM的高粱叶片病斑图像自动分割提取方法研究  

A Method of Sorghum Leaf Disease for Image Automatic Segmentation and Extraction Based on the SVM

作　　者：白文斌[1] 白帆[1] 贺文文[1] 王伟仁[1] 程彦俊[1] 刘璋[1] 

机构地区：[1]山西省农业科学院高粱研究所,山西晋中030600

出　　处：《农学学报》2014年第6期101-106,共6页Journal of Agriculture

基　　金：山西省农业科学院科技攻关项目"高粱叶部病害智能识别;分析及决策系统"(YGG0716);山西省科技攻关项目"高粱种质资源创制及专用新品种选育--北方粒用高粱高效控高技术研究"(20130311003-2);山西省农业科学院重点项目"粒用高粱节水高产高效农机农艺一体化技术研究"(2013ZD02)

摘　　要：为实现高粱叶片病斑的自动化无损监测,利用支持向量机(SVM)技术对高粱叶片病斑图像进行自动分割提取研究。结果表明,通过选取RGB、HIS和Lab 3种颜色空间的颜色特征值可以消除对作物病斑拍照时产生的光照、亮度等影响。在MATLAB软件环境下调用LIBSVM软件对病斑图片中的病斑图像像素点和背景图像像素点建立支持向量机分类模型,可以实现对病斑的高效分割和高质量提取。分割提取效果与人眼识别的病斑图像高度吻合。如果利用大量采集的病斑图像进行模型训练,就可以真正实现完全自动化的病斑分割、提取和判别。因此,该研究对建立完全自动化的作物病斑图像识别系统意义重大。In order to achieve the automated nondestructive monitoring of sorghum leaf ulsease spot, the author uses support vector machine （SVM） technology to research automatic segmentation and extraction of sorghum leaf disease spot image. The results showed that selecting the color feature values of the 3 kinds of color spaces （RGB, Ills and Lab） could eliminate the influence of the light brightness when you took a photo. In the MATLAB software environment using LIBSVM software to establish support vector machine （SVM） classification model of disease spot image pixels and background image pixels, could implement disease spot image efficient segmentation and high-quality extraction. The disease spot image which was extracted automatically by programs could closely match the recognition by the human eye. If using a large number of sampled disease spot image to train model, could achieve the disease spot image fully automated segmentation, extraction and determination. So this research has very important significance to build fully automated crop disease spot image recognition system.

关 键 词：支持向量机 LIBSVM 高粱 病害 图像分割 自动识别 

分 类 号：S431[农业科学—农业昆虫与害虫防治]