Map of Europe
In Spring of 2018 the project has been extended. The extension aims at implementation of the S2GLC classification approach for classification of the whole Europe. As a source of training data two existing LC databases are used: Corine LC and High Resolution Layers.
Due to the characteristics of the European landscape and LC databases used, the global way of proceeding has been modified. Intensive tests were carried out in selected test areas spread across Europe in different climate zones. Applications of different sets of spectral channels, classification features and sources of training data were tested as well as the method for selecting images for classification from a time series and possibility for detecting new LC classes.
Due to the amount of data the whole classification process has been performed using the CREODIAS infrastructure. Over 15 000 Sentinel-2 images from the year 2017 are classified fully automatically with the use of algorithms and software developed by CBK PAN. Validation of the results has been performed by IABG mbH
Land Cover Map of Europe
Automatic approach requires automated training data selection. For this purpose we have used Corine Land Cover and Copernicus High Resolution Layers and two spectral indices.
Software and computations
Computation are performed with CREODIAS platform. Software has been developed by CBK PAN.
Land cover maps for test sites:
The total number of tiles randomly selected for validation is 55. Number of samples per tile is 800, what gives 44 000 reference samples in total.
Accuracy for the whole Europe with 13 land cover classes resulted in 86% overall accuracy:
Accuracy for the whole Europe with 10 land cover classes resulted in 89% overall accuracy:
Accuracy assessment per country: