Morphological Leaf Classification of Species of The Genus Licania Occurring In The Amazon Using Embedded Spiking Neural Networks

This paper presents a study on the morphological leaf classification of species of the genus Licania (Chrysobalanaceae) occurring in the Amazon, integrating botanical analysis methods with contemporary Artificial Intelligence–based techniques. Species of this genus present ecological relevance, pharmacological potential, and recurrent use in traditional Amazonian medicine, making accurate taxonomic identification an essential factor for scientific and ethnobotanical applications. The proposed approach employs Spiking Neural Networks (SNNs) for the recognition of morphological patterns in leaf images, considering characteristics such as shape, venation, and leaf dimensions. Inspired by the functioning of biological neurons, SNNs exhibit intrinsic temporal processing capability and high computational efficiency, making them suitable for taxonomic identification tasks based on complex data. The proposed model was trained and evaluated using images obtained from different sources, encompassing natural variations in acquisition conditions and environmental factors. The results indicate high discriminative capability of the model, with consistent separation among the analyzed species, even in scenarios of morphological similarity. This study contributes to advances in plant taxonomy by demonstrating the potential of Spiking Neural Networks as an auxiliary tool for automated species identification, particularly in contexts of high morphological plasticity, such as tropical environments.