Performance Analysis of Different Optimization Techniques on CNN in Hyperspectral Images of Pichavaram Mangrove Region

Mangroves are vital coastal ecosystems that provide essential services including wildlife habitat, carbon sequestration and more. The Pichavaram mangrove in Tamilnadu highlights these benefits through its role in coastal protection and biodiversity conservation. Accurate analysis of this area is essential for effective environmental monitoring and conservation. Remote sensing technologies, especially Hyperspectral Imaging (HSI) capture detailed spectral information across numerous wavelengths. When combined with Deep Learning (DL) particularly Convolutional Neural Network (CNN), HSI offers significant benefits for mangrove analysis. This study evaluates the impact of various optimization techniques such as Stochastic Gradient Descent (SGD), Root Mean Square Propagation (RMSProp), Adaptive Gradient (AdaGrad), AdaDelta (Adaptive Delta) and Adaptive Moment estimation (Adam) on CNN performance using HSI datasets like EO1 Hyperion and AVIRIS-NG from Pichavaram mangroves. Among these, the Adam optimizer consistently outperforms the others, proving to be the most effective for HSI mangrove classification. This analysis highlights the necessity of choosing the appropriate optimization technique to maximize the potential of DL models in environmental monitoring and future conservation efforts in mangrove ecosystems.