About the research

With the exponential increase in computer processing power, data acquisition rates, data complexity and data volume, traditional methods of, e.g., analyzing, storing and transferring the ever increasing amounts of data are becoming inadequate for many fields, including remote sensing (RS). Big Data is a term that is often used to describe this phenomenon. Recently, a set of methods called Deep Learning (DL) have shown big potential in analyzing and extracting valuable information from Big Data. These methods are based on deep neural networks and today there is tremendous growth in different implementations, architectures and capabilities of those methods.

The goals of this project are to develop powerful DL methods for two important fields of RS. Firstly, to develop methods for the problem of image fusion in RS, where multispectral or hyperspectral images of low spatial resolution are fused with a high spatial resolution image of single or more channels to obtain an image with high spectral and spatial resolution. Secondly, to develop methods for supervised and unsupervised classification of hyperspectral images, where the focus will be primarily on unsupervised classification. Both goals represent valuable contributions to the respective fields of RS. Image fusion is a key technology in products such as Google maps and land cover classification is among the most important fields within the field of RS.

Keywords: Deep Learning, Big Data, Remote Sensing, Image Fusion, Classification

See further details