ANALYZING BIG REMOTE SENSING DATA VIA SYMBOLIC MACHINE LEARNING
Martino Pesaresi, Vasileios Syrris, Julea Andreea Maria · Joint Research Centre (European Commission) · 2016
This paper discusses the applicability of a recently introduced classification method for Earth Observation data and justifies its eligibility in the framework of Big Data processing and analysis. The main novelty that this classifier brings forward in the Remote Sensing domain is the automated learning from symbolic representations as opposed to the traditional quantitative approach. The method has been designed to work as a melting pot at which, data arriving from diverse sources and with heterogeneous characteristics (sensors, scales, time periods) are fused and refined to valuable information. A description of the classification method is presented briefly. In addition, results from the production of the multi-temporal medium resolution Global Human Settlement Layer that derived from Landsat satellite imagery of the last 40 years are demonstrated and assessed.