| In this thesis, a new method is presented for extracting stand-level forest inventory parameters from high spatial resolution satellite imagery based on analysis of image objects. High spatial resolution satellite imagery is expected to replace aerial photographs as the primary data source for forest inventory applications; however, conventional digital image analysis techniques, based on the analysis of individual image pixels, have been unable to provide the quality of information required for most forest inventory purposes. The object-based analysis approach used in this thesis, which differs from conventional pixel-based approaches, was successful in extracting forest inventory information from Ikonos satellite imagery. An image segmentation routine was established to partition the Ikonos data into image objects representing forest stand components. Decision trees were used to identify empirical relationships between standard forest inventory parameters and image object metrics from a high-dimensional data set of spectral- and spatial-based features derived mainly from the Ikonos image data. |
