Abstract
This paper examines the effects of multiple view angles on the classification of forward-modeled, high-quality, multispectral reflectances in a Canadian boreal forest region using a decision tree classifier (C4.5). Bidirectional reflectance factors (BRFs) from the seven-band moderate resolution imaging spectroradiometer (MODIS) are reproduced from high-quality composite model parameter datasets that were retrieved using a daily rolling version of an operational algorithm developed for direct broadcast and that were successfully used in earlier research. To assess the classification accuracies, we adopted descriptive and statistically rigorous techniques based on a confusion matrix and a 10-fold cross-validation method. The results show that the classification accuracies derived from the modeled MODIS BRFs in the principal plane are not substantially different, with the exception of a few directions, relative to bi-hemispherical reflectances (the white sky albedo) in the MODIS bidirectional reflectance distribution function (BRDF) Albedo product. The highest and lowest overall classification accuracies are those acquired by the seven-band Nadir BRDF-Adjusted Reflectances (approx. 77.745% ± 3.036) and the seven-band hotspot reflectances (approx. 72.18% ± 2.27). Analysis of per-class accuracies of eight land cover classes with different structures shows that the herb class and the broadleaf dense class have high per-class accuracies (mostly greater than 90%) from various view angles; whereas, other classes have relatively low per-class accuracies that tend to change with the view zenith angle and that are somewhat higher in the close-to-nadir and backward directions than in the forward scattering directions. Further investigation reveals that the classification accuracies derived from the reproduced MODIS BRFs are negatively correlated with the within-class variances of these BRF input features. Moreover, such correlations are higher in backward scattering directions (including the nadir direction) than in forward scattering directions. In summary, the effects of multiple view angles on the classification of MODIS BRFs reproduced from the MODIS BRDF model using a decision tree classifier (C4.5) are mainly related to anisotropic variance patterns of the BRFs in the principal plane.
Talk to us
Join us for a 30 min session where you can share your feedback and ask us any queries you have
Disclaimer: All third-party content on this website/platform is and will remain the property of their respective owners and is provided on "as is" basis without any warranties, express or implied. Use of third-party content does not indicate any affiliation, sponsorship with or endorsement by them. Any references to third-party content is to identify the corresponding services and shall be considered fair use under The CopyrightLaw.