Semi-supervised hyperspectral image classification using a new (soft) sparse multinomial logistic regression model

Abstract

In this work, we propose a new semi-supervised classification algorithm for remotely sensed hyperspectral images. The main contribution of this work is the development of new soft sparse multinomial logistic regression (S <sup xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">2</sup> MLR) model which exploits both hard and soft labels. In our terminology, these labels respectively correspond to labeled and unlabeled training samples. In order to obtain the soft labels, we use a recently proposed subspace-based MLR algorithm (MLRsub). The proposed semi-supervised algorithm represents an innovative contribution with regards to conventional semi-supervised learning algorithms that only assign hard labels to unlabeled samples. The effectiveness of our proposed method is evaluated via experiments with a widely used hyperspectral image collected by the Airborne Visible Infra-Red Imaging Spectrometer (AVIRIS) over the Indian Pines region in Indiana. Our results indicate that the proposed method provides state-of-the-art performance when compared to other methods.