摘要
针对遥感图像分类问题提出了一种基于遗传算法和K近邻的SVM决策树方法。算法以基于类分布的类间分离性测度为准则,利用遗传算法对传统的SVM决策树进行优化,生成最优(较优)决策树。在分类阶段,对容易分的节点利用SVM进行分类,而对可分离性差的节点采用SVM和K近邻相结合的分类方法,最终实现多类别分类。实验结果表明,与传统的分类方法相比,该算法的实验效果较好,可有效地提高遥感图像的分类精度。
This paper presented a SVM decision-tree algorithm based on GA and KNN. First, GA was used to create optimal or near-optimal decision-tree, which defined a novel separability measure. Then in the class phase, standard SVM was used to make binary classification for the divisible nodes, and SVM combined with KNN were used to classify the fallible nodes. Finally, achieved the multi-classification by the SVM decision-tree. Experimental results show that the proposed method can effectively improve the classification precision of remote sensing image in comparison to traditional classification methods.
出处
《计算机应用研究》
CSCD
北大核心
2012年第3期1146-1148,1151,共4页
Application Research of Computers
基金
辽宁省科技计划资助项目(2010401010)
关键词
遗传算法
K近邻
支持向量机决策树
遥感图像分类
genetic algorithm
K-nearest neighbors
support vector machine （ SVM ） decision-tree
classification of remote sensing image
