Authors: Andong Ma*, Texas A&M University, Anthony M. Filippi, Texas A&M University, Zhengcong Yin, Texas A&M University
Topics: Remote Sensing, Geographic Information Science and Systems, Land Use
Keywords: Hyperspectral image classification, deep learning, recurrent neural networks, spatial contextual information
Session Type: Paper
Start / End Time: 12:40 PM / 2:20 PM
Room: Virginia B, Marriott, Lobby Level
Presentation File: No File Uploaded
Hyperspectral remote-sensing images (HSI) can entail both abundant spectral and spatial information, which generally provides the enhanced capability of distinguishing different objects from one another, relative to multispectral images, and play an important role in a variety of geography domains. Classification is a common objective when analyzing hyperspectral images, where each pixel is assigned to a predefined label. Deep learning-based algorithms have been introduced in the remote-sensing community successfully in the past decade and have achieved significant performance improvements compared with conventional models. In this research, a novel recurrent neural network (RNN)-based framework for HSI classification is proposed. Within this classification scheme, unlabeled data that have been widely investigated in semi-supervised learning is employed efficiently for a classification task, as unlabeled data can also provide significant information, in conjunction with the limited labeled data. In addition, spatial contextual information, which has already been utilized to reduce the “salt-and-pepper” phenomenon in remote-sensing/HSI classification, is also investigated in our classification framework. We conduct experiments on two benchmark HSIs, and three conventional supervised classifiers are utilized for comparison. The experimental results demonstrate that the proposed methods achieve marked improvements in classification performance relative to the other state-of-the-art methods considered, and the “salt-and-pepper” phenomenon is alleviated significantly.