沙化土地类型图像识别方法对比研究——以毛乌素沙地典型地区为例

Comparative Study of Image Identification Methods for Deserted Land Type——A Case Study of Mu Us Sandy Land

以毛乌素沙地典型地区为例,以CBERS-2影像作为基本数据源,分别使用光谱知识法和波谱角分类法对研究区流动沙地、半固定沙地、固定沙地和沙化耕地进行信息提取,探索和比较了两种不同方法针对不同沙化土地类型的分类模型和精度。结果表明,使用光谱知识法对各类型沙化土地分类总体精度为76.04%,使用SAM方法分类的总体精度为83.21%,说明对于CBERS影像中的沙化土地而言,SAM方法能够较为快速准确地提取各类型沙化土地信息。

Spectral knowledge method and spectral angle mapper（SAM）are used to extract the desertification information,including flow sand,semi-fixed sand,fixed sand,and sandy plantation by taking Mu Us sandy land as a typical area and CBERS-2 as the source of data.The extracted precision of the different methods are analyzed and compared.Results show that the summarized precision of spectral knowledge method is 76.04% and SAM is 83.21%.This indicates that the extracted precision of SAM is higher than spectral knowledge method for extraction of the desertification information on CBERS images.