Crop type classification in the federal state Brandenburg using Machine Learning models and multitemporal, multispectral Sentinel-2 imagery

Abstract

Regarding environmental changes and more extreme weather conditions, forecasting crop yields and capturing crop conditions is more crucial than ever and is also vital in managing remaining resources and, therefore, securing the food supply. Against this background, this work will develop a technology that can collect real-world data on different cultivations from space resulting in a processor that can classify different crops in the area of Brandenburg. The functionality of the developed processor is focused to the classification of five different crops with the utilization of remote sensing and machine learning methods. Furthermore, the processor will be used to draw conclusions from phenological developments in different areas of the federal state of Germany Brandenburg. The work will reveal challenges when classifying crops from space and makes proposals on how to solve them.