针对序列投影寻踪聚类模型用于多视极化SAR图像分类存在的速度慢的问题,文章提出了一种基于Bootstrap方法的快速投影寻踪实现方法。文中定义了拟最佳投影方向,给出了用于分类的Bootstrap样本选取准则,于是,只需要较少的样本就可寻找到拟最佳投影方向,实现极化SAR图像的快速分类。分类结果和原有的序列投影寻踪聚类分类结果一样,与直接采用极化散射特性的方法相比,都能极大地改进分类结果。同时相比于原有算法又极大地提高算法实现的速度。
Aim.Ref.3,our previous paper,presented the SPPCM method for classifying Pol-SAR(polarimetric synthetic aperture radar) images.We believe that it is good but unfortunately it is too slow.We now improve it by using Bootstrap method.Section 1 of the full paper defines the quasi-optimal projection pursuit direction given in eq.(4),and then derives the Bootstrap sample selection criterion shown in eq.(7),which requires only a few Bootstrap samples.Section 2 presents existing classification methods based on scatt...