基于邻域直方图的玉米田绿色植物图像分割方法  

Image Segmentation of Green Plants in Corn Field Based on Neighborhood Histogram

作　　者：邵乔林[1] 安秋[1] 

机构地区：[1]南京农业大学工学院,江苏南京210031

出　　处：《江西农业学报》2011年第5期126-128,135,共4页Acta Agriculturae Jiangxi

基　　金：南京农业大学青年科技创新基金项目(KJ09030)

摘　　要：为了准确分割绿色植物与土壤背景,克服田间环境的复杂性,比如植物阴影、残渣的存在和光照强度变化等外界因素带来的影响,采用图像处理和支持向量机方法,以超绿特征(2G-R-B)灰度图像为研究对象,以像素点邻域组成的灰度直方图特征作为输入特征向量,通过试验选出最佳训练参数和最优邻域窗口模型。实验表明,该方法适应不同光照强度,并且可以减小噪声、植物阴影和残渣对图像分割带来的影响,得到完整的分割图像。In order to accurately segment green plants and soil background,to overcome the complexity of field environment(such as the shadow and residues of plants,the changes of light intensity),the methods of image treatment and support vector machine were used to study the gray image with ultra green feature(2G-R-B).The gray histogram characters being composed of the neighborhood pixels were taken as the input eigenvectors,and the best training parameters and the optimized neighborhood window model were screened o...

关 键 词：邻域直方图 图像分割 支持向量机 机器视觉 

分 类 号：TP391.41[自动化与计算机技术—计算机应用技术]