基于稀疏成分分析的遥感影像分类

Remote sensing image classification based on sparse component analysis

遥感影像分类一直是遥感研究的重点、难点和热点之一。针对经典的主成分分析法在不同地物的光谱存在重叠相关时,分类效果欠佳的缺陷,提出一种基于稀疏成分分析的遥感影像分类法。该方法利用稀疏性提取源信号,不要求源成分之间互不相关。实验结果表明,与主成分分析方法相比,基于稀疏成分分析的分类结果更可靠、更准确。

The classification of remote sensing images is a key issue and focused subject in remote sensing image processing. Considering that the classification result of classical principle component analysis （PCA） is not satisfying when the spectra of different ground objects are related, a new classification method based on sparse component analysis （SCA） is presented. The proposed method utilizes the sparseness characteristic to extract source signals, and does not demand the sources be independent. The experimental result shows that compared to principle component analysis, the classification result of SCA is more reliable and more accurate.