二次样本筛选的高光谱图像分类研究

College of Information and Communication Engineering Harbin Engineering University

主动学习能够在有标记样本较少的分类任务中得到较好的分类结果,其中熵值装袋算法最为常用,其利用熵值来衡量样本的不确定性,但熵值并不能完全地代表样本的不确定度。针对这一问题,本文提出二次样本筛选的分类算法,通过超像素分割进行边缘区域样本筛选,选择出不确定度较高的样本。利用熵值装袋算法对区域筛选样本进行二次筛选,选择信息量较高的样本。实验表明,该方法可以得到更理想的分类效果。

Active learning can get better classification results in the classification task with fewer labeled samples,and entropy bagging algorithm is the most commonly used.It uses entropy to measure the uncertainty of samples,but entropy can not completely represent the uncertainty of samples.In order to solve this problem,this paper proposes a classification algorithm of secondary sample selection,which selects the samples with high uncertainty by super-pixel segmentation.The entropy bagging algorithm is used to screen the samples selected in the previous step for the second time,and the samples with higher information content are selected.The experiments show that this method can obtain better classification effect.