Abstract: Hyperspectral Images have hundreds of spectral bands, which makes them rich in information but challenging to classify. Deep learning has been proved to be useful for hyperspectral image classification, but it is important to choose the right architecture and to train the model on a big and diverse dataset. This project offers a deep learning method for classifying hyperspectral images. The proposed method applies preprocessing techniques including noise reduction and feature extraction to extract spectral-spatial characteristics from hyperspectral images using a convolutional neural network (CNN) architecture. Using supervised learning, a sizable dataset of hyperspectral images is used to train the CNN. Then, new hyperspectral images are classified into various land cover classes using the learned model. Additionally, a Support Vector Machine (SVM) is employed alongside the CNN to enhance the classification accuracy. The SVM acts as a complementary classifier, taking the features extracted by the CNN to improve the decision boundaries between different classes. The integration of SVM with CNN helps in achieving better generalization and robustness in classification results. The results of the proposed system show that the suggested technique performs more accurately for hyperspectral image classification using the combined approach of convolutional neural network and support vector machine.
Keywords: Convolutional Neural Network, Hyperspectral, Neural Network, Deep Learning, Spectral, Support Vector Machine.
Copyright Form