Hedgerow Mapping using Convolutional Neural Networks

In this study, the feasibility of high spatial resolution optical imagery for the mapping of hedgerow objects is assessed. Two different Convolutional Neural Networks are applied to IKONOS imagery based on in situ data provided by the Bayrisches Landesamt für Umwelt (LfU). While overall results are promising, in a next step the combined use of areial images with Sentinel-2 time series should be tested. The generated hedge maps are supposed to support biotop mapping and monitoring activities conducted at federal state level.