NOISY BAND SELECTION BASED ON THE INTEGRATION OF THE STACKED-AUTOENCODER AND CONVOLUTIONAL NEURAL NETWORK APPROACHES FOR HYPERSPECTRAL DATA

Seleção de bandas ruidosas baseada na integração de Stacked-AutoEncoder e redes neurais convulacionais para dados hiperespectrais

Autores

DOI:

https://doi.org/10.5016/geociencias.v42i2.16976

Resumo

RESUMO - A presença de ruído em imagens hiperespectrais causa degradação e dificulta a eficiência no processamento para a classificação da cobertura terrestre. Portanto, a remoção do ruído ou a detecção automática de bandas ruidosas em imagens hiperespectrais torna-se um desafio para pesquisas na área de sensoriamento remoto. Para enfrentar esse problema, um modelo integrado (SAE-1DCNN) é apresentado nesse estudo, baseado nos algoritmos de Deep Learning conhecidos como: Stacked-Autoencoders (SAE) e Redes Neurais Convolucionais (CNN) para a seleção e exclusão de bandas ruidosas. O modelo proposto emprega as camadas convolucionais para melhorar o desempenho dos Autoencoders focados na discriminação dos dados de treinamento por meio da análise da assinatura hiperespectral do pixel. No contexto do SAE-1DCNN, as informações são comprimidas e a informação redundante é identificada e removida. Isso é possível dado à eficiência da arquitetura profunda baseada em camadas convolucionais e de agrupamento. Os resultados obtidos demonstram a capacidade do modelo em identificar automaticamente bandas ruidosas, sugerindo que a nossa abordagem tem potencial e pode representar uma alternativa promissora para a detecção de bandas ruidosas no pré-processamento de dados hiperespectrais.

Palavras-chave: Bandas ruidosas. Seleção de atributos. Redes neurais convolucionais. Stacked-autoencoders. Dados hiperespectrais.

 

ABSTRACT - The presence of noise on hyperspectral images causes degradation and hinders efficiency of processing for land cover classification. In this sense, removing noise or detecting noisy bands automatically on hyperspectral images becomes a challenge for research in remote sensing. To cope this problem, an integrated model (SAE-1DCNN) is presented in this study, based on Stacked-Autoencoders (SAE) and Convolutional Neural Networks (CNN) algorithms for the selection and exclusion of noisy bands. The proposed model employs convolutional layers to improve the performance of autoencoders focused on discriminating the training data by analyzing the hyperspectral signature of the pixel. Thus, in the SAE-1DCNN model, information can be compressed, and then redundant information can be detected and extracted by taking advantage of the efficiency of the deep architecture based on the convolutional and pooling layers. Hyperspectral data from the AVIRIS (Airborne Visible/Infrared Imaging Spectrometer) sensor were used to evaluate the performance of the proposed automatic method based on feature selection. The results showed effectiveness to identify noisy bands automatically, suggesting that the proposed methodology was found to be promising and can be an alternative to identify noisy bands within the scope of hyperspectral data pre-processing.

Keywords: Noisy bands. Feature selection. Convolutional neural network.  Stacked-autoencoders. Hyperspectral data.

Biografia do Autor

Mario Ernesto JIJÓN-PALMA, Universidade Federal do Paraná

Departamento de Geomática, Programa de Pós-graduação em Ciências Geodésicas, Avenida Coronel Francisco Heráclito dos Santos, 210 – Jardim das Américas – Curitiba, Paraná, Brasil.

Ernesto Jijón-Palma graduated in geographical and environmental engineering from Army Polytechnic School (ESPE-2008, Universidad de las Fuerzas Armadas). He holds his master’s degree and PhD in geodetic sciences from Federal University of Paraná. Currently, he is doing a post doctorate in computer science at the Federal Technological University of Paraná. Research area is hyperspectral and multispectral remote sensing, image processing and, photogrammetry.

Caisse AMISSE, Universidade Rovuma: Nampula, Nampula, MZ

Caisse Amisse graduated in oceanography from Universidade Eduardo Mondlane in 2008. He holds his master’s degree and PhD in geodetic sciences from Federal University of Paraná and currently he is a professor at Rovuma University in Nampula Mozambique–surveying concentration remote sensing 

Jaime Carlos MACUÁCUA, Universidade Eduardo Mondlane

Universidade Eduardo Mondlane.

Avenida Julius Nyerere, 3453, Maputo - Moçambique

Jorge Antonio Silva CENTENO, Universidade Federal do Paraná

Jorge Antonio Silva Centeno graduated in civil engineering from the Federal University of Mato Grosso do Sul in 1988, received a master of science in water resources and environmental sanitation from the Federal University of Rio Grande do Sul in 1991 and a PhD in geodesy at the University of Karlsruhe in 2000. Currently, he is a professor at the Federal University of Paraná. He has experience in remote sensing, image processing, and photogrammetry.

Downloads

Publicado

2023-09-15

Edição

Seção

Artigos