Harnessing Ensemble Learning: Land Cover Classification using Hyper Spectral Imaging

Hyperspectral imaging (HSI) is a powerful technology for land cover classification due to its ability to capture rich spectral information across hundreds of narrow bands. However, the high dimensionality and inherent noise in HIS data pose significant challenges for accurate and efficient classification. This study proposes a robust framework that enhances land cover mapping by integrating dimensionality reduction techniques – Principal Component Analysis (PCA) and Linear Discriminant Analysis (LDA) – with machine learning classifiers including Support Vector Machine (SVM), Random Forest, K-Nearest Neighbors (KNN), and Gaussian Naïve Bayes. To further improve classification accuracy and generalization, an ensemble model is employed, leveraging the strengths of individual classifiers. The framework is evaluated on the widely used Indian Pines Dataset, demonstrating improved classification performance and computational efficiency. Preprocessing steps effectively reduce noise and redundancy in the data, ensuring better feature representation. The results show that the proposed approach yields reliable land cover classification, making it suitable for real-world applications in agriculture, environmental monitoring, and urban planning. This study highlights the importance of combining dimensionality reduction with machine learning to address the limitations of high-dimensional hyperspectral data.