基于TM影像的面向对象地表覆被信息提取

为验证基于TM影像的面向对象分类方法对复杂地区地表覆被信息提取的可行性,以地处西南地区的渝北为例进行实验。利用样本数据对各个波段的光谱特征进行分析,取得对各波段覆被探测能力的初步认识;基于光谱特征的多尺度分割,运用面向对象分类方法对其分类。面向对象的分类方法总精度和Kappa系数分别为88.42%和0.854 7,将其与监督、非监督分类结果对比分析。结果表明,该方法有效抑制了"椒盐"现象,取得较好的分类结果。

Land Cover Information Extraction Based on Object-oriented Classification

Taking Yubei District which located in the southwest of China as object,based on TM remote sensing images,we obtained the cognition ability for each band of land cover through the information quantity statistics and spectral characteristic analysis.Using object-oriented classification of multi-resolution segmentation extracted the information of land cover.The total classification accuracy was 88.42% and the Kappa coefficient was 0.854 7.Then the result was compared with the results of supervised and unsupervised classification,the accuracy of object-oriented classification was higher than supervised and unsupervised classification.It approves that the technology of object-oriented classification of multi-resolution segmentation is applicable to the classification of remote sensing image of rolling country.