基于不同分辨率遥感影像的分类方法对比研究

Study on Classification Methods Based on Different Resolution Remote Sensing Image

基于4种不同分辨率的遥感影像数据,分别为30 m分辨率的Landsat-8数据,融合Landsat-8多光谱波段和全色波段的15 m分辨率数据,5.8 m分辨率的资源3号卫星数据以及融合后2.1 m分辨率的资源3号卫星数据。采用ISO-DATA、最大似然分类法和面向对象分类法对影像进行分类,对分类方法的效果以及分辨率变化对面向对象分类方法的精度影响进行分析。结果显示在低分辨率影像中,面向对象方法受到限制,分类效果相比传统方法没有太大改善;而在高分辨率影像中,面向对象方法分类效果很好,并且随着分辨率提高分类精度也相应的提高。

This paper is based on four different resolutions remote-sensing images.30 m resolution Landsat-8 image data, 15 m resolution image fused by multispectral bands and panchromatic band of landsat-8 data, 5.8 m resolution ZY-3 image data and 15 m resolution fusion image.ISO-DATA , maximum likelihood classification and object-oriented image classification method were used.The effects of different classification methods and impact of resolution changes on the accuracy of object-oriented method were analyzed.The results show that object-oriented approach is limited in low resolution images, the classification result is not much im-proved compared to the traditional classification method.While in the high -resolution images the object -oriented classification method works well, and classification accuracy is improved along with resolution increase.