Publication

Abstract : Due to the benefits of metaheuristic optimization techniques, in this paper, we introduced flower pollination optimization algorithms for finding optimal bands in airborne hyperspectral images and classification using a modified wavelet Gabor filter (MGFNet) convolutional neural network. The proposed flower pollination optimization algorithm has been investigated to select optimal bands from hyperspectral images with deep wavelet features, which are nonlinear, discriminant, and invariant. These bands are effective for hyperspectral image classification, object detection. Moreover, in demand to maintain the widespread issue of imbalanced samples for the classification of Hyperspectral image, we have selected only optimized bands rather than whole bands of hyperspectral images. More importantly, we proposed a modified Gabor based wavelet filter which helps to extract exact information from the spatial and spectral features of hyperspectral imagery. The proposed approaches are carried out on three hyperspectral datasets: Indian pines, Pavia University, and Salinas scene hyperspectral images. In addition, the proposed FPO and MGFNet open up a new frame for further research.