Mapping 'Lantana camara': Accuracy Comparison of Various Fusion Techniques

Abstract

Fusion of panchromatic and multi-spectral QuickBird satellite imagery was carried out to evaluate the impact of fusion techniques on classification accuracies for Lantana mapping. This study compared four image fusion techniques, namely Brovey, Hue-Saturation-Value, Principal Components, and Gram-Schmidt Spectral Sharpening. Classification accuracy assessment was calculated using an error matrix for all images. Gram-Schmidt and Principal Components spectral sharpening techniques had an overall accuracy of 90.5 percent and 89.5 percent and a kappa coefficient of 0.85 and 0.84, respectively, compared to the MS image which had an overall accuracy of 86.3 percent and a kappa coefficient of 0.79. Brovey transformation and HSV performed poorly in the supervised classification with overall accuracies of 64.2 percent and 76.8 percent and kappa coefficients of 0.48 and 0.65, respectively. Visual and statistical analyses of the fused images showed that Gram-Schmidt and Principal Components spectral sharpening techniques preserved spectral quality much better than the Brovey and HSV fused images.