|Wu, J - UNIVERSITY OF MINNESOTA|
|Bauer, Marv - UNIVERSITY OF MINNESOTA|
|Manson, Steve - UNIVERSITY OF MINNESOTA|
Submitted to: International Society for Photogrammetry and Remote Sensing Proc
Publication Type: Peer Reviewed Journal
Publication Acceptance Date: August 4, 2007
Publication Date: March 6, 2008
Citation: Wu, J., M. Bauer, D. Wang, and S.M. Manson. 2008. A comparison of illumination geometry-based methods for topographic correction of QuickBird images of an undulant area. ISPRS Journal of Photogrammetry and Remote Sensing 63: 223-236. Interpretive Summary: Evaluating spatial variability at meter or sub-meter resolutions is made possible with aerial or new satellite platforms such as QuickBird from Digital Globe. This new capability generates interested in applications including precision agriculture on a field level. The limitation with the increased resolution is that the effect of topography-induced illumination variations becomes more of concern even in moderately sloped areas. The objective of this study was to analyze topographic effects on QuickBird images of an undulant area under different illumination and ground conditions. Four illumination geometry-based topographic correction methods were evaluated. Results showed that empirical correction was the most effective method for all spectral bands in both solar and view directions, while the cosine correction gave the worst results. Mixed results were found with the C correction or the Minnaert correction methods. None of these methods significantly changed the spatial variability of spectral radiances. The overall conclusion is that application of high spatial imageries tends to be more sensitive to ground-level elevation changes than data from low or medium resolution platforms such Landsat or MODIS, so in-situ geometric correction would increase the accuracy of the spectral assessment.
Technical Abstract: The high spatial resolution of QuickBird satellite images makes it possible to show spatial variability at fine details. However, the effect of topography-induced illumination variations become more evident, even in moderately sloped areas. Based on a high resolution (1 m) digital elevation model generated with high-frequency real-time kinematic global position system measurements, this study assessed topographic effects on QuickBird images of an undulant area (with a maximum slope of 7.4°) under different illumination and ground conditions. For land surfaces that were characterized by a non-Lambertian reflection, significant bidirectional variations in spectral radiances were found in all bands. The effectiveness of four illumination geometry-based topographic correction methods was evaluated. The results indicated that the empirical correction was the most effective method for all spectral bands in both solar and view directions, while the cosine correction gave the worst results. The C correction (in the solar direction) and the Minnaert correction reduced topographic effects, but not as effectively as the empirical correction. For the Lambertian, topographic effects were substantial only in the near infrared band in the solar direction. Bidirectional variations of spectral radiances in other bands and/or view directions were minimal and topographic corrections may not be necessary. None of these methods significantly changed the spatial variability of spectral radiances, although the histogram distributions were greatly modified by the cosine correction and the Minnaert correction.