摘要
文章针对城市遥感图像的目标分布特点,提出一种基于改进DTSVM的遥感图像分割方法。实验引入样本的聚类特性改善DTSVM模型分类精度,对城市遥感图像中的区域进行语义标注并提取特征,通过训练改进分类模型得到分割结果。实验结果表明,该方法能比较准确地分割出关注语义的目标区域,并有效避免了遥感图像的过分割问题。
In view of the object distribution characteristics of urban remote sensing images, this paper proposes a remote sensing image segmentation method based on improved decision tree support vector machine（DTSVM）. The clustering characteristics of testing samples is used to improve the classification accuracy of DTSVM model, and features are extracted from the semantieally-annotated regions on urban remote sensing images. Then by training the improved DTSVM model with these features, the segmentation results of testing images are obtained. The experimental results demonstrate that the proposed method provides a satisfactory segmentation of concerned semantic objects and the over-segmentation of remote sensing images is effectively avoided.
出处
《合肥工业大学学报（自然科学版）》
CAS
CSCD
北大核心
2011年第3期383-386,共4页
Journal of Hefei University of Technology：Natural Science
关键词
遥感图像分割
语义
支持向量机
remote sensing image segmentation
semantics
support vector machine（SVM）
