Deep 2D Convolutional Neural Network Architecture for Hyperspectral Land Cover Classification: A Comparative Study with KNN

Abstract

In recent years, deep learning techniques have received a great deal of attention in the context of hyperspectralimage (HSI) classification, particularly with regard to land cover mapping. Although 2D convolutionalneural networks (CNNs) are now widely used in this field, this study presents a refined, deeply structured2D-CNN architecture that is specifically designed for spatial–spectral integration. Rather than introducinga novel concept, the contribution lies in the balanced design of the architecture, which integrates dropoutand batch normalisation to enhance accuracy and generalisability on benchmark datasets. The proposednetwork includes 10 convolutional layers organized into three blocks, each followed by max-pooling, batchnormalization, and dropout layers to reduce overfitting and improve model robustness. A fully connectedclassifier with Softmax activation performs the final prediction. We trained the architecture using the SalinasValley dataset, which contains 54,129 labeled pixels across 16 land cover classes. The data weremeticulously segmented into two distinct components: the initial segment encompassed the primary dataset, while the subsequent segment comprised the ensuing data. It is noteworthy that 70% of the data wasallocated for training purposes. The remaining 30% of the budget was allocated for testing purposes. Thetraining was executed for 100 epochs by employing the Adam optimizer and categorical cross-entropy lossfunction. The 2D-CNN model demonstrated superior performance in terms of classification accuracy whencompared with the KNN approach. The 2D-CNN model attained a classification accuracy of 94%, whilethe KNN method achieved 88%. The findings indicate the efficacy of deep 2D-CNNs (Convolutional NeuralNetworks) in the classification of hyperspectral land cover. The results also demonstrate the networks’suitability for implementation in large-scale remote sensing projects.