MACHINE LEARNING IN REMOTE SENSING

Kurzfassung

Remote Sensing plays an essential role in Earth Observation and thus in understanding of the complex relationship between bio/geo-physical processes and human welfare. The growing quality and quantity of remotely sensed data make manual interpretation infeasible and require accurate and efficient methods to automatically analyse the acquired data. This chapter discusses two Machine Learning approaches that address these tasks on the example application of semantic segmentation of polarimetric Synthetic Aperture Radar images. It shows the importance of proper classifier design and training as well as the benefits of automatically learned features.