<![CDATA[This study aims to classify urban and green land images by employing the Singular Value Decomposition (SVD) method for dimensionality reduction and the K-Nearest Neighbors (KNN) algorithm for classification. The dataset used consists of 600 images (300 urban and 300 green land) with a resolution of 256x256 pixels, sourced from the Kaggle "Aerial Landscape Images" dataset. Each image was transformed into a feature vector, then reduced using SVD, where this study compares the use of 5 components and 20 components. The dataset was subsequently divided into 80% training data and 20% testing data for classification using KNN with k=3. Performance evaluation was conducted via confusion matrix and the calculation of accuracy, precision, recall, and F1-score. The results showed that the model with 5 SVD components achieved the highest accuracy of 95.00%, outperforming the 20-component model (91.67%). This finding demonstrates the effectiveness of SVD-KNN, but shows that a higher number of components can degrade performance. The limitation of the purely color-based method was also identified during testing on "residential area" images, which possess overlapping features between “Urban” and “Green Spaces”.]]>
Copyrights © 2026
