Evaluation and Comparison of the Performance of Water Body Extraction Methods from Remote Sensing Data with Emphasis on Traditional Methods, Machine Learning and Deep Learning Approaches
DOI:
https://doi.org/10.68924/wbh2vj53Keywords:
Methods, Method accuracy, Remote sensing, Satellite images, Water area extractionAbstract
Water forms an important part of human life, and without water, the survival of living beings on Earth is impossible. Water resources include oceans, seas, wells, etc. Research and identification of these important resources are of great value to human health. Water plays a very important role in the political, social, and economic aspects of countries. This research aims to review methods for extracting water from satellite images and introduce the best method .In this research, the modified normalized difference water index (MNDWI) methods , Water Ratio Index (WRI) ،Normalized Difference Water Index (NDWI), (SMDPSO+LFE) ،(Faster R-CNN) Automated Water Extraction Index (AWEI) ,Support Vector Machine (SVM), Random Forest (RF) ,NDWI-PCs ،Water Index (WI) ,Feature Classification with Deep Learning (DLFC), Local Radiative Energy Peak (LREP) ,Double Band Normalized Difference Water Index (DBNDWI) ،Linear Discriminant Analysis (LDA) ,Artificial Neural Network (ANN)، (DeepLab-v3+ ،( LinkNet ،Multi-scale Feature Extraction Network (MSFEN), Water Extraction Surface Temperature Index (WESTI)، Multilayer Perceptron (MLP) ،Corrected Normalized Difference Water Index (CNDWI) ،Decision Tree (DT) and the convolutional neural network (CNN) method have been investigated, Each method has its own advantages and disadvantages. After examining the methods, the results of the Faster R-CNN method have better performance than other methods. This method has excellent performance on two datasets of Sentinel-2 and Landsat-8 (OLI) satellite images for extracting water areas, the results show that this method has higher accuracy compared to other methods and can identify water more effectivelyand accurately.
Downloads
Published
Versions
- 2026-11-07 (2)
- 2026-10-31 (1)