Estudo da extração de bordas de reservatório utilizando múltiplas técnicas de fusão de imagens

Abstract

The diverse techniques of remote sensing support the monitoring of water reservoirs and especially in the extraction of its edges. Among them, highlight techniques known as image fusion and digital filtering. The application of this methodology consists of combining different spectral and spatial characteristics to generate an image with better spatial resolution and then apply a directional filter. There are several types of image fusion and digital filtering and among them will be highlighted in this research the IHS (intensity, hue and saturation), Principal Components (CP), Brovey and Gram-Schmidt fusions and the Sobel directional filter. The present study developed a comparison between these techniques for the extraction of reservoir contour with greater detail. The study was done in images of CBERS-4 satellite, MUX and Panchromatic 5 meters for the reservoir Engenheiro Francisco Saboia-Ibimirim, PE-Brazil. The comparison between the results showed that, although similar, CP, Brovey and Gram-Schmidt present more consistent results when compared to the IHS. We also performed a study based on the Euclidean distance calculation, comparing the edge generated by the segmentation by regions growth with the edges generated from the Sobel filter application