一种改进的SVM遥感图像分类方法  

An improved SVM algorithm for satellite image classification

作　　者：李蕾 LI Lei(Shaanxi Surveying and Mapping Production Supervision and Inspection Station,Ministry of Natural Resources,Xi’an Shaanxi 710054,China)

机构地区：[1]自然资源部陕西测绘产品质量监督检验站,陕西西安710054

出　　处：《北京测绘》2023年第6期903-907,共5页Beijing Surveying and Mapping

摘　　要：支持向量机(SVM)是一种新兴的机器学习算法,常用在遥感影像分类研究中。针对样本数目不均衡时标准SVM算法的分类结果精度不佳的问题,本文根据不同类别的样本数对样本定权,提出了基于样本数加权的SVM算法。采用GeoEye卫星的高分辨率遥感影像对该算法进行验证,相应结果和标准SVM算法进行对比。结果表明,训练样本数较多时加权SVM算法与标准SVM算法均取得较好效果,但当训练样本数不均衡时加权SVM算法可有效补偿其不利影响,精度远优于标准SVM算法。Support vector machine(SVM)is an emerging machine learning algorithm,which is commonly used in satellite image classification research.To address the problem that the classification results of standard SVM are not accurate while the number of samples is unbalanced,this paper proposed a weighted SVM algorithm and the samples were weighted according to the number of samples in different categories.This algorithm was validated using high-resolution GeoEye satellite images,and the corresponding validated results were compared with those of the standard SVM algorithm.The results showed that both the weighted SVM algorithm and the standard SVM algorithm achieved accurate results when the number of training samples was large.But when the number of training samples was unbalanced,the weighted SVM algorithm could effectively compensate for its adverse effects,and the accuracy of the weighted SVM algorithm was much better than that of the standard SVM algorithm.

关 键 词：加权支持向量机 遥感图像分类 不均衡 分类精度 

分 类 号：P237[天文地球—摄影测量与遥感]