The factor of scale in remote sensing

Abstract

A method that measures the spatial structure of images as a function of spatial resolution is presented for selecting the appropriate scale for remote sensing. Graphs are obtained by imaging the scene at fine resolution and then collapsing the image to successively coarser resolutions while calculating the local variance. For the spatial resolution of SPOT and TM imagery, local image variance is relatively high for forested and urban/suburban environments, indicating that information-extraction techniques using texture, context, and mixture modeling are appropriate for these sensor systems. For agricultural environments where local variance is low, more traditional classifiers are appropriate.