基于面向对象的多端元光谱混合分析方法

A Method of Multiple Endmember Spectral Mixture Analysis Based on Object Oriented

为了充分利用中等分辨率遥感影像的空间、纹理和光谱信息,解决影像中存在的大量混合像元问题,提高多端元光谱混合分析方法的效率,本文提出了基于面向对象的多端元光谱混合分析方法。首先利用面向对象的分析方法对研究区遥感影像进行分类,得到分类结果图;然后对分类结果图中包含多种土地覆盖类型的对象进行多端元光谱混合分析,获得各土地覆盖类别的丰度图及分解残差图;最后对基于面向对象的多端元光谱混合分析结果与多端元光谱混合分析结果进行对比分析。结果表明,基于面向对象的多端元光谱混合分析方法能够得到比较连续且和实际地物分布较为吻合的分析结果,其分解精度和效率优于多端元光谱混合分析方法。该方法可更有效地提取地表覆盖信息,为研究区域性生态环境变化、模拟分析提供了有效的信息提取方法。

In order to make full use of the medium spatial resolution remote sensing images＂ space, texture and spectral information, to solve the problems of mixed pixels widely spreaded in images and to improve the efficiency of multiple endmember spectral mixture analysis method, the method of the multiple endmember spectral mixture analysis based on object oriented was proposed in the paper. First, we classified remote sensing image of the study area using the object oriented analysis method, and obtained the classification results map. Secondly, we obtained the abundance map of each land cover category and decomposition of residual plot by the method of multiple endmember spectral mixture ana- lyse. Finally, we analyzed the results of multiple endmember spectral mixture analysis and multiple endmember spec- tral mixture analysis based on object oriented. The results showed that the method of the multiple endmember spectral mixture analysis based on object oriented could obtain the classification results which were more continuous and more coincided with the actual distribution feature and with higher accuracy and decomposing efficiency. The method can be more effective in land cover information extraction, and provide effective information extraction method for the study of the regional ecological environment change and simulation analysis.