Object-based analysis of Ikonos-2 imagery for extraction of forest inventory parameters

A method is presented for deriving forest inventory information from Ikonos-2 imagery based on the analysis of image objects rather than more conventional pixel-based image analysis approaches. For a 77 km(2) study area in southwestern Alberta, Canada, image objects representing homogeneous landscape components were delineated from Ikonos-2 data using an image segmentation routine. Decision tree statistical analyses were used to identify correlations between metrics derived from spectral and spatial properties of the image objects and field-derived samples of individual forest inventory parameters. The strongest relationships were observed for classes of discrete land-cover types, species composition, and crown closure.