ارزیابی قابلیت فیوژن طیفی و مکانی در بهبود دقت نقشه پوشش اراضی
Mohammad Mansourmoghaddam, Iman Rousta, Hamid Reza Ghafarian Malamiri, Mohammad Hossein Mokhtari
Most remote sensing satellites, such as Landsat, Spot, and QuickBird, produce many types of images, such as Multi-spectral (MS) and Panchromatic. Through the combination of high spatial and high spectral image characteristics, image fusion may generate a high spatial and high spectral resolution MS Image. The ability of SAR Sentinel-2, optical Landsat-8, and their two types of fused images for Yazd, Iran, to improve the accuracy of the Land Cover (LC) map using supervised Maximum Likelihood Classification (MLC) was examined and quantified in this study. The present study, using the Gram-Schmidt Pen Sharpening fusion method, first focused on the spatial fusion of Landsat-8 images with Sentinel-2 images. Then the Sentinel spectral bands were also fused with Landsat’s fused spectral bands and formed a new series of data. Finally, 4 data series of Landsat-8 30-m images (IM1), Sentinel-2 10-m images (IM2), spatial fused image (FIM1), and spectral-spatial fused image (FIM2) were classified using the MLC method and were evaluated. Also, the Normalized Optimum Index Factor (NOIF) index was developed based on the Optimum Index Factor (OIF) index and the number and combination of bands with the desired amount of information were examined. Results have shown that, because Landsat-8 has a higher spectral resolution than Sentinel-2 and Sentinel-2 has a higher spatial resolution than Landsat-8, their combination has boosted the information of images used for classification. As a result, the ideal NOIF values have been defined as 0.6 to 1. Furthermore, as compared to the initial unfused image of Landsat-8, the combination might improve overall classification accuracy by 10% and the Kappa coefficient by 16.5%. Also, the list of ideal band combinations with NOIF greater than 0.6 has been reported to aid researchers in Yazd in doing their categorization more properly and quickly.