Studi Perbandingan Performansi Metode Clustering Berbasis Fitur Citra Multispektral untuk Klasifikasi Tutupan Lahan pada Citra Satelit Resolusi Tinggi

Rosanti Rosanti (Universitas PGRI Pontianak)
Amalia Shifa Aldila (Universitas Jakarta Internasional)
Gregorio Antonny Bani (Universitas Aryasatya Deo Muri)

Polygon: Jurnal Ilmu Komputer dan Ilmu Pengetahuan Alam
Vol. 2 No. 2 (2024): Maret: Polygon : Jurnal Ilmu Komputer dan Ilmu Pengetahuan Alam

Land cover classification using multispectral satellite imagery plays a vital role in environmental monitoring, spatial planning, and sustainable resource management, although the high spectral dimensionality and spatial complexity of such data require an effective clustering method to generate accurate segmentation. This study aims to compare the performance of multispectral spectral feature-based clustering algorithms in high-resolution land cover classification. Multispectral features including reflectance bands, NDVI, NDBI, and NDWI were extracted and normalized prior to implementing three algorithms—K-Means, DBSCAN, and Agglomerative Clustering—where the optimal number of clusters was determined using the Elbow method and Silhouette analysis. Performance evaluation using internal validation metrics showed that K-Means achieved the highest Silhouette Score () and the lowest Davies-Bouldin Index (DBI) (), indicating superior intra-cluster compactness and inter-cluster separation compared to both DBSCAN and Agglomerative Clustering. These findings conclude that centroid-based clustering is more effective for handling structured spectral distributions in high-resolution multispectral imagery.