摘要
运用SPOT-5全色和多光谱影像,采用支持向量机(SVM)法对森林植被进行分类研究,探讨了SVM法的分类能力以及纹理信息在森林植被分类中的影响。结果表明:Gram-Schmidt光谱锐化法是北京山区SPOT-5影像最佳的融合方法;SVM法在高分辨率影像森林植被分类中精度较高,不同核函数对分类精度的影响不显著;基于灰度共生矩阵产生的纹理信息能够提高SVM法的分类精度,3×3窗口是提高分类精度的最佳纹理窗口。
The experiment was conducted to classify the forest vegetation with support vector machine （SVM） method based on SPOT-5 panchromatic and multispectral images and explore the ability with SVM method and the effect by texture informa-tion in forest vegetation classification .Gram-Schmidt spectral sharpening method is the best fusion method forSPOT -5 im-age in Beijing mountain areas.SVM method has higher classification accuracy with the fine resolution images in the forest vegetation extraction.There is no significant difference on classification accuracy with different kernel functions.Image texture information from Gray level co-occurrence matrix （ GLCM） method can improve the classification accuracy by SVM method, and the best texturewindow of ×3 windows can improve the classification accuracy obviously.
出处
《东北林业大学学报》
CAS
CSCD
北大核心
2014年第1期51-56,共6页
Journal of Northeast Forestry University
基金
国家"863"计划课题(2012AA102001-5)
教育部高等学校博士学科点专项科研基金资助课题(20100014110002)
关键词
影像融合
Gram-Schmidt光谱锐化法
灰度共生矩阵
支持向量机
植被分类
Image fusion
Gram-Schmidt spectral sharpening method
Gray level co-occurrence matrix （GLCM）
Support vector machine （SVM）
Vegetation extraction
