Evaluating the accuracy of manual classification in satellite images using supervised algorithms

This research evaluated manual land-cover classification using images extracted from Sentinel-2. Supervised algorithms were applied to validate and enhance this process. Three algorithms were selected based on their computational efficiency: K-Nearest Neighbors (KNN), Random Forest (RF), and Support Vector Machine (SVM).The results show that KNN achieved optimal performance, demonstrating a strong balance between accuracy, F1 score, and runtime compared to RF. RF, in turn, obtained higher precision and F1 scores, indicating its superior ability to correctly identify classes; however, it required greater computational resources. SVM exhibited lower performance in the evaluated metrics but achieved a shorter runtime. Nevertheless, it was identified as the algorithm with the greatest limitations in separating classes within this dataset derived from the different study areas.Overall, the comparison confirmed that the manual classifications developed in QGIS are supported and validated by the application of these supervised methods. The use of such algorithms contributes to improving accuracy, consistency, and efficiency in geospatial classification tasks.