An evaluation of empirical image-based topographic correction models on a Landsat ETM+ image of rural Nevada, USA.

摘要

Remotely sensed images acquired over rugged terrain show illumination distortions resulting from differences in incident sun angles relative to terrain angles. This can result in areas with steeper illuminated slopes facing away from the sun receiving less radiance than those facing towards the sun. The relationship between radiance and incident angle can be modelled in order to reduce these distortions. Empirical topographic normalisation methods based on this relationship were applied to a Landsat-7 ETM+ image of a rural area of Nevada, USA. All methods were image-based, incorporating DEM-derived variables. Image-based methods do not require knowledge of scene area landcover types, but they do require a suitable DEM. B-correction and VECA normalisation methods were employed, along with a proposed modification to the VECA technique. As expected, for anisotropic reflection (generally for visible light), those models based on a linear correlation between radiance and the incidence angle performed best. For more isotropic reflection (mainly infrared), those methods employing the relationship between the logarithm of the radiance and the incidence angle, produced generally better results.