Image data fusion for flood plain mapping

Abstract

Mini multispectral sensors are an insufficient source of imagery to perform appropriate analyses, due to the low spatial resolution of imagery. Therefore, in many cases it is necessary to conduct data fusion. However, using common pansharpening methods may not be sufficient. Therefore, the authors propose a new method for processing and sharpening multispectral data acquired with a mini multispectral camera, for mapping purposes, especially in flooded areas and flood plains, in rapid time. The proposed algorithm of sharpening is based on the use of the decorrelation process of multispectral images and their transformation to the YCBCR color space, where the luminance component is converted to the blue band of the high resolution image and then an inverse transformation to the RGB color space is performed, resulting in imagery with a high spatial and spectral resolution. The results of our research were compared with pansharpened multispectral images generated using classical methods: IHS, PCA, Brovey, Ehlers, wavelet and multiplicative transforms. Moreover, an assessment of a quality of the sharpened spectral images by determining: R-RMSE, ERGAS and Q Index indicators was performed.