Abstract
Object-based semi-automated segmentation and classification approaches have gained importance in the analysis of remote sensing data over the last few years. Particularly when it comes to operational processing of multi-seasonal input data, independent and robust algorithms are needed. At the German Aerospace Center (DLR) a new method for forest type classification has been developed, covering all processing steps for object-based classification. An automatic adaptation of scene-specific feature values for the classification is implemented, based on automated extraction of feasible ground data. Therefore, no manual sampling of training data is necessary. For classification of mixed forests on the basis of IKONOS data, a special algorithm was developed that can be adapted to any kind of mixed forest definition. Forest age classes are derived based on a digital surface model. The developed method can be used for area-wide forest-type classification on the basis of high and very high-resolution satellite data.
Acknowledgements
The authors would like to thank Eberhard Tschach from LANU for his substantial support, the reviewers for their valuable comments, and the ESA for funding the GSE Forest Monitoring project.