超图自动学习与最优聚类框架结合的波段选择

Band selection combining hypergraph automatic learning with optimal clustering framework

图像中标记样本的缺乏和复杂的噪声使得从高光谱图像(HSI)中选择判别波段成为一项具有挑战性的任务。提出了一种超图自动学习与最优聚类相结合的波段选择方法。将整个频带空间随机划分为几个不同维度的子空间(LV),每个子空间由一组相关频带组成的训练样本的低维表示组成;将所有子空间的样本投影到标签空间后,利用超图自动模型保留这些投影的局部流形结构,保证同一类的样本距离较小,并使用一致性矩阵整合不同子空间对应的波段;通过聚类公式得到排序顺序,选出最优的聚类结果。在3个高光谱图像数据集上的实验表明,此方法与其他方法相比更具有竞争力。

The lack of labeled samples and the complex noise in the images make the selection of discriminative bands from hyperspectral images(HSI)a challenging task.A band selection method combining hypergraph automatic learning with optimal clustering is proposed.The whole band space is randomly divided into several subspaces(LVs)of different dimensions,each subspace consists of a set of low-dimensional representations of training samples consisting of its associated bands.After the samples of all subspaces are projected into the label space,the hypergraph automatic model is used to preserve the local manifold structure of these projections to ensure that the samples of the same class have a small distance.The consistency matrix is used to integrate the bands corresponding to different subspaces.The ranking order is obtained through the clustering formula,and the optimal clustering result is selected.The classification experiments on three HSI data sets show that our method is compared with other comparison methods.