Classificação de Imagens Hiperespectrais utilizando Redes Neurais Convolucionais para Caracterização da Ocupação Desordenada do Solo sobre um Recorte do Parque Nacional da Tijuca, RJ

A novel approach using Convolutional Neural Networks (CNNs) and deep learning algorithms achieved high accuracy—0.92 with data augmentation—in classifying hyperspectral images to map land cover, including disordered urban occupation in the Tijuca National Park (PNT), a critical Atlantic Forest fragment near Rio de Janeiro. This methodology effectively distinguished between forest and urban areas, demonstrating its potential for monitoring challenging terrains like PNT, which provides vital environmental services such as erosion prevention and carbon capture. The study highlights CNNs' efficacy in pattern recognition for land-use mapping, offering a scalable solution for sustainable development and conservation efforts.

Do Vale, I. G., Delgado, A. R. S., & Silva, R. M. (2022). Classificação de Imagens Hiperespectrais utilizando Redes Neurais Convolucionais para Caracterização da Ocupação Desordenada do Solo sobre um Recorte do Parque Nacional da Tijuca, RJ. Revista Brasileira de Geografia Física, 15(1), 561–571. https://doi.org/10.26848/rbgf.v15.1.p546-557