Document Type : Research Paper
Author
Assistant Professor of Reserch Institute of Forest and Rangelands, Agricultural Research, Education and Extension Organization (AREEO), Tehran, Iran
10.22034/ijf.2023.328404.1849
Abstract
The purpose of this study is to investigate the performance of different algorithms in increasing the accuracy of land cover map using sentinel images. The study area is a sheet (1.25000) with an area of 15782.6 hectares. Land cover map using Mahalanobis distance, maximum likelihood, minimum distance, neural network, parallelepiped, support vector machine, spectral angle mapper, spectral information divergence and binary encoding algorithms using the best band composition It was performed from 12 bands (2, 3, 4, 5, 6, 7, 8, 8a, 11, 12, NDVI, SAVI) and training area obtained from field information and satellite images. 70% of the samples were used for classification and 30% for evaluating the accuracy the classified maps. The results show that the three classifications of support vector machine, neural network and maximum likelihood have the highest accuracy. The support vector machine classifier is much more accurate than the other two classifier with a very small difference. It should be noted that to improve the classification, support vector machine kernels (linear, polynomial, radial basis function and sigmoid) and spectral angle mapper settings (6 modes) and parallelepiped (2 modes) were used. The results show that the support vector machine classifier by the 6th degree polynomial function method has the highest accuracy. Then the map was Postprocessed with the highest accuracy and the overall accuracy was 96.63 and the kappa coefficient was 0.9393. Examination of 21 classification modes on sentinel images in this study shows a complete review of classification algorithms in comparison with the studies performed.
Keywords
Main Subjects
)