One Class SVM for buildings detection on Sentinel-2 images

Author
Michał Krupiński, Stanisław Lewiński – Centrum Badań Kosmicznych PAN
Summary
Senitnel-2 satellites, as a part of European initiative - Copernicus - provide unique opportunity for Earth surface monitoring with high spatial, spectral and temporal resolution. Urban areas mapping is one of the domains where this kind of data may be found very useful. In this research, usability of Sentinel-2 data for buildings detection will be analysed using One-class SVM algorithm. For selected area of interest, number of issues have been analysed, including automatic training data preparation, the optimal number of training points, date of acquisition and algorithm parameters.