DEEP LEARNING APPROACH FOR REMOTE SENSING IMAGE ANALYSIS

The paper explores how multimedia approaches used in image understanding tasks could be adapted for remote sensing image analysis. In a first step, we show on 3 channels color images through the UC Merced Land Use Dataset how Deep Learning approach provides a significant performance increase compared to Bag of VisualWords approach. In a second step, we propose an extension of deep learning scheme to deal with hyperspectral data. The proposed scheme is based on a 3D architecture which jointly processes spectral and spatial information.