混合像元分解的理论与方法-以神木县为例

识别分类是遥感技术应用中的重要一环,而混合像元是影响识别分类精度的主要因素之一。为了提高识别分类精度,本文提出了四种混合像元分解方法,它们是匹配像元分解、折半像元分解、双邻像元分解、相关像元分解。混合像元分解前,计算机分类识别精度仅为63.65％,混合像元分解后,其识别分类精度达88.67％,可见混合像元分解的效果是很显著的。

The Theroeies and Methods of Mixed Pixel Decomposition

Recognition and classification is an important link for the applications of remote sensing techmologies, and the existence of mixed pixels is one of the main factors affecting the preci sion of recognition and classification. In order to improve the precision, four approaches are put forward in this paper, including matching pixel decomposition, dimidiate pixel decomposition, two neighbouring pixels decomposition, relative pixel decomposition. The precision isonly 63.65 percent before the decomposition, while the precision reaches 88.87 percent after that. From this point, the effect of the mixed pixel decomposition is remarkable.