Optimized Hyperspectral Crop Classification using Coot Bird Algorithm

Hyperspectral remote sensing (HSRS), also known as imaging spectroscopy, is a cutting-edge method for capturing spectral data across numerous spatial locations in a wide array of narrow spectral wavelengths, enabling the generation of detailed and coherent images. HSRS involves collecting digital images across contiguous spectral bands, spanning the visible, Near Infrared (NIR), Thermal Infrared (TIR), and MidInfrared (MIR) regions of the electromagnetic spectrum. For agricultural applications, remote sensing techniques facilitate continuous and quantitative monitoring, providing crucial physical and chemical information about vegetation. This study introduces the Agricultural Crop Classification using Coot Bird Optimization with EntropyBased Feature Fusion (ACC-CBOEFF) technique for hyperspectral imagery (HSI). The ACC-CBOEFF approach leverages feature fusion and parameter tuning to classify various crop types effectively. It incorporates Gabor filtering as a preprocessing step and employs entropy-based feature fusion using ShuffleNet, EfficientNet, and DenseNet models for robust feature extraction. Additionally, the Deep Variational Autoencoder (DVAE) system is utilized for crop type classification, with the Coot Bird Optimization (CBO) method optimizing DVAE hyperparameters. Simulations on benchmark datasets demonstrate the superior performance of the ACC-CBOEFF technique compared to other state-of-theart methods