Individual tree crown (ITC) delineation is an important step in facilitating aspects of sustainable forest management. This is due to the valuable information that can be derived from the individual tree level: biomass measurements, species identification, gap analysis, and carbon content. In Indonesia, individual tree detection is done by field surveys which are expensive and time consuming, while ITC delineation for small-scale monitoring purposes are normally done by visual interpretations using satellite imagery. In this study, two different methods were proposed in order to assess the potential of OBIA techniques to delineate ITCs in the tropical forest in Central Kalimantan, Indonesia: a seed based region growing algorithm and the Thiessen polygon method. To do so, we first detected the tree peak locations in all the three study areas (SA01, SA02 and SA03) using a method based on variable smoothing with DSM height retaining in the local maxima algorithm. Thereafter, we used these tree peak locations as tree seeds to perform the seed based region grow algorithm in eCognition and to construct the basic shape of the tree crown segmentation in the Thiessen polygon method. For both ITC delineation methods, the DSM layer and RGB orthorectified imagery were used as the main input layers to produce the delineated ITC objects. The results showed that the accuracy of these methods were comparable with results from other studies conducted in the same forest type. Based on the results from the accuracy assessment, the region growing method achieved a total accuracy of 61.39%, which is significantly higher than the accuracy derived from the Thiessen polygon method (55.88%). Further work is needed to improve the accuracy of the ITC delineation by using different types of data such as: hyperspectral imagery and terrestrial laser scanning ground measurements.
Download:
Publication year
2015
Authors
Language
English
Keywords
trees, crown, identification, peat, swamp, tropical forests
Geographic
Indonesia
