Burnt area mapping with aerial data

The aim of this thesis is to leverage aerial data for the automatic mapping of areas burnt by wildfires. The data were captured by a Light Manned Aircraft (LMA) during 3 separate flights over Delfi, Domokos, Yliki and Prodromos. The obtained imagery comprises of Green, Red, Red-Edge and Near-Infared channels and has a spatial resolution of 60cm.

In order to use the LMA data in a mapping algorithm, it needs to be cleaned, processed and cropped into smaller patches. Subsequently, the student can choose among a number of Machine Learning or Deep Learning methods and train them on the data in order to obtain an accurate perimeter of the burn scar. The model(s) will also be evaluated using appropriate performance metrics.

Supervisor: Maria Sdraka