Land cover information is vital data about many surface processes. As a representation of an observable Earth surface, land cover maps have vital importance for geo-scientists, land managers and government. Thus, updating land cover maps should be regular routine.
The main aim of this work is to create a software application for semi-automatic land cover classification based on geographic object-based environment.
Two different classification methods were implemented to this software application. A sample-based classification and an index-based classification characterize core methods to classify land cover classes. In the index-based classification method, a combination of several vegetation indices, spatial data and object features were examined and tested. Conducted accuracy assessment results showed that both methods can achieve satisfying classification results in all three levels of classification of LCCS Dichotomous phase. The highest average accuracy of 98.66% was received in first level classification. The lowest accuracy was received by sample-based classification, due to misclassification of spectrally similar land cover objects. This issue was solved by the index-based classification.
The main result of this master thesis was the software application for land cover classification. An index-based classification was set up for this study. Results showed comparatively high accuracy. Additionally, it can be further developed to become a robust approach for perform land cover classification.