一种基于 KFCM的 SVM遥感图像机场目标分类算法

在遥感图像机场目标分类方面,支持向量机(SVM)有着广泛的应用,但由于样本不平衡问题以及不确定性数据的存在,传统SVM算法的分类精度与效果还无法令人满意。为提高传统SVM分类器的性能,文章将建立在模糊理论基础上的模糊核C-均值聚类算法(KFCM)用于处理遥感数据的不确定性问题,并通过聚类分析后的目标子图,剔除非目标样本的同时保留了目标样本,较好地解决了样本不平衡问题。将基于KFCM的SVM分类算法用于遥感图像机场目标的分类,实验结果和性能分析表明该算法分类性能优于传统SVM算法。

An Airport Target Classification Algorithm in Remote Sensing Image Based on KFCM and SVM

Support vector machine （SVM） has wide application in remote sensing image airport target classifica- tion, but the traditional SVM algorithm can not get satisfying result in respect of classification accuracy and ef- fect due to sample imbalance and uncertainty data in RS image. To improve the performance of traditional SVM algorithm, the Fuzzy kernel C-means clustering algorithm（KFCM） established in fuzzy theory was used to solve uncertainty problem in RS image, and target subgraph obtained by clustering analysis was used to get rid of nontarget sample and retain target sample, then researched a SVM classification algorithm based on KFCM. Fi- nally a airport classification experiment was done on remote sensing image, experiment results and performance analysis showed that the proposed algorithm had more superior performance than the traditional SVM algorithm.