基于光谱空间密度分析的边缘提取

Edge Extraction Based on Density Analysis in Spectral Space

针对传统的边缘提取方法大部分不适应高光谱数据的特点 ,提出了基于光谱空间密度分析边缘提取的思想。在分组主分量变换提取第一主分量作为特征维的基础上 ,采用面向对象的二次判别边缘的方法 ,通过立体判决将光谱空间中低密度超椭球体集群视为真实边缘点集群。试验表明 ,此方法是合理可行的。

The density analysis of super dimensional spectral space for edge extraction is presented. On the basis of the first principle component after grouping PCA, the fundamental methods of object oriented two times edges determination is proposed. Every edge point called candidate detecting from each first component will perform classification in feature space as vectors. Edge feature points are distributed in the form of lower density hyperellipsoid in spectral space because they have class characteristics like common points of large areas. Real edge points are considered as those in lower density zones of spectral space.