Document Type : Research Paper
Authors
1 Department of Physical Geography, Faculty of Social Sciences, University of Mohaghegh Ardabili, Ardabil, Iran.
2 Department of Physical Geography, Faculty of Social Sciences, University of Mohaghegh Ardabili, Ardabil, Iran
3 Department of Physical Geography, Faculty of Social Sciences, University of Mohaghegh Ardabili, Ardabil
Abstract
Cereals are considered one of the most important sources of dietary protein, and wheat is a significant cereal crop with high protein content. Currently, the rapid and excessive population growth and the perceived shortage of available resources to meet essential human needs are among the biggest challenges facing the world. Accurate and up-to-date statistics and information on agricultural capacities form the foundation of proper planning and management in agricultural affairs.
Methods: In this study, Sentinel2-L2A satellite images were initially downloaded, and the Normalized Difference Vegetation Index (NDVI) was extracted using the set of images containing ground reflectance data. Then, the Support Vector Machine (SVM) and Random Forest classification algorithms were applied to the images using the R programming language in the Jupyter Notebook environment.
Results: Finally, it was observed that the Random Forest algorithm performed better and more appropriately, with an overall accuracy of 93% and a kappa coefficient of 87%, compared to the Support Vector Machine algorithm, which had an overall accuracy of 90% and a kappa coefficient of 82%. This preference is due to its higher accuracy and kappa coefficient, indicating a greater agreement with reality and higher prediction accuracy.
Conclusions: The results of these algorithms showed that each algorithm has its own strengths and weaknesses. The Support Vector Machine algorithm is used in many classification problems due to its simple structure and adequate performance. However, in this study, it performed weaker compared to the other algorithm, the Random Forest. The Random Forest algorithm usually provides accurate results due to its ability to combine different models and reduce the effect of overfitting. Nevertheless, its high computational complexity can be problematic in larger applications.
Keywords
Main Subjects
)