中文摘要：

针对传统的特征提取方法都是基于向量模型,导致处理时维数极高,且极易丢失像素空间信息的问题,该文将二维主成分分析引入高光谱影像特征提取领域,该方法在保持影像原有空间结构信息的前提下,通过多变量线性变换,求取最佳投影方向.下仅能提高同类地物的聚团性、避免分类后地物混淆,还能消除最终分类结果的“麻点”现象,在试验中验证了有效性.

英文摘要：

Traditional feature extraction method are based on vector model, which not only leads to high dimensions, but also loses pixel＇s space information. In the paper, a feature extraction based on two dimensions principal component analysis was proposed to solve the problems. This method could make full use of data＇s space information, it could also find the best project＇s direction by linear transformation. At the same time, it could elimate the final results of the ＂pitting＂ phenomena and avoid classification surface features confusion. The experiment result showed that the the proposed method was effective.