Object-oriented Classification of Remote Sensing Data for the Identification of Tree Species Composition

This paper deals with the classification of tree species composition from Ikonos imagery (4m resolution) based on the object-oriented image analysis in eCogniton software. The image was acquired over a man-planted forest area with proportion of various forest types (conifers, broadleaved, mixed) in the Krušné Hory Mts., Czech Republic. In order to enlarge class feature space, additional channels were produced by low-pass filtering, IHS transformation and influence of various Haralick texture measures on classification was also examined. The principal component calculated from original Ikonos bands was applied in the pre-processing phase. The segmentation and classification were conducted on three levels to be incorporated into the hierarchical image object network. The higher level separated image into smaller parts regarding the stand maturity and structure, the lower (detailed) level assigned individual tree clusters into classes for the main forest species. The third level was created to distinguish forest/non-forest boundaries. Classification accuracy was assessed by comparing the automated technique with the field inventory data using Kappa coefficient. Apart from transferable rule base creation, the study aimed at determining appropriate scale for species composition estimation using common image data. Therefore the methodology will be tested on colour aerial photos in the further research.