面向对象的高分辨率影像分类与信息提取

采用面向对象遥感影像分类方法对高分辨率遥感影像进行了信息提取实验,并将其与基于像元方法的信息提取结果进行了对比分析。实验研究表明,在目视效果上,传统方法的分类结果图中“椒盐现象”非常明显,而面向对象方法可以有效地避免“椒盐现象”；在分类精度上,面向对象方法分类结果的总体精度、Kappa系数、生产者精度、用户精度、Hellden精度和Short精度均明显高于传统方法,各类地物提取效果显著提高,总分类精度提高21.76%,Kappa系数提高0.2756。面向对象方法在高分辨率遥感影像信息提取中具有明显的优势。

Object-oriented Classification and Information Extraction Based on High Spatial Resolution Remote Sensing Image

The object-oriented classification method was used to extract land use/land cover information from high spatial resolution remote sensing images,and the results are compared with the traditional information extraction method.It showed that the accuracy was greatly improved,and the method can avoid salt and pepper noise effectively.The classification results are easier to understand and interpretative.Overall precision,Kappa,producer precision,user precision,Hellden precision and Short precision of object-oriented classification are better than that of traditional classification method.The extraction effect of each ground object increased significantly.overall precision increased 21.76% and Kappa coefficient increased 0.2756.These experiments show that the object-oriented method is superior to traditional method in high resolution remote sensing information extraction.