Recent improvements in sensor technology have led to an increased availability of remote sensing images with high spectral and spatial resolution. Spectral and spatial analysis of information provided by remote sensing images makes it possible to identify and classify different land-covers of an observed area of interest.
In particular, VHR multispectral images are characterized by a geometric resolution in the order or smaller than one meter. This type of data allows to precisely characterizing the geometrical properties of the different objects (e.g., buildings, streets, agriculture fields, etc.) in the scene under investigation. This remote sensed data provide very useful information for several applications related to the monitoring of the natural environment and of human activities (urban environment).
Hyperspectral images are composed by hundreds of contiguous spectral channels, covering a wide spectral range of frequencies, in which each pixel is a highly detailed representation of the reflectance of the materials presented on the ground. Hyperspectral remote sensing is an emerging, multidisciplinary field with diverse applications that are based on the principles of imaging spectrometry, and hyperspectral data processing. The increasing quality of hyperspectral images, which contain a vast amount of information regarding the land-coverage, provides high quality and efficiency Earth’s surface monitoring applications, such as forestry.
The aim of the research is to develop new advanced automatic systems for very high resolution (VHR) and hyper- multi- spectral remote sensing images classification.
Classification in remote sensing is a complex task that employs a number of processes aiming at addressing the challenging issues that emerge from the nature of the images, requiring knowledge in different topics, e.g., machine learning, pattern recognition, mathematical morphology, image segmentation, feature extraction and selection.