Enhancement of image classification with image sharpening and neural networks

The aim of the study is to make image segmentation processes using artificial neural networks on two satellite images in the Seyhan Basin area, one unprocessed and the other with sharpening, and to create a comparison table on the efficiency of the sharpening process from the analysis of the results. The study starts by obtaining the satellite images of the Seyhan Basin from the Sentinel-2 mission, producing a new image from these images by sharpening then separating them into two images in total. Then, ROIs were created on these data and two different labeled image sets were obtained. These sets of images were then trained in an artificial intelligence algorithm and two different training models were produced. With the help of these, the first two images were classified. Accuracy analysis with ground truth data was performed on these. The obtained accuracy results were compared with each other based on general and class, and a comparison table was obtained that revealed the difference in efficiency between the two images.