A regularized multi-metric active learning framework for hyperspectral image classification

Abstract

Utilization of both spectral and spatial features for hyperspectral image classification can often improve the classification accuracy. However, the high dimensionality of the input data and the limited number of labeled samples are two key challenges for supervised techniques. In this paper, a regularized multi-metric learning approach is proposed for feature extraction and combined with active learning (AL) to deal with these issues simultaneously. In particular, distinct metrics are assigned to different types of features and then learned jointly. Also, the proposed regularizer helps to avoid overfitting at early AL stages by taking advantage of the unlabeled data information. Finally, multiple feature are projected into a common feature space, in which a new batch-mode AL strategy combining uncertainty and diversity is performed in conjunction with k-nearest neighbor (ANN) classification to enrich the set of labeled samples. Experiments on a benchmark hyperspectral dataset illustrate the effectiveness of the proposed framework.