Classification of Remote Sensing Scenes Based on Neural Architecture Search Network

Abstract

Remote sensing image scene classification plays an important role in remote sensing image retrieval, land-use identification and urban planning. Deep learning brings great opportunity to the research in this field, but it transfers the difficulty of traditional characteristic engineering to the design of network structure. In this paper, we focus on the automatic design of the network model and propose a remote sensing scene classification method based on Neural Architecture Search Network (NASNet). We further use the transfer learning technology to make the designed network well migrated to the remote sensing scene classification data set. This method can automatically build the appropriate network structure according to the application. We compare the proposed method on a publicly large-scale dataset with several convolutional neural network (CNN) models. The experimental results demonstrate that the proposed method provides state-of-the-art performance compared with the traditional artificial neural network.