Banca de DEFESA: Daniele Dias Sousa

Uma banca de DEFESA de MESTRADO foi cadastrada pelo programa.
STUDENT : Daniele Dias Sousa
DATE: 31/01/2023
TIME: 09:00
LOCAL: Virtual
TITLE:

Methods for Reconstruction of Hyperspectral Images for the Brazilian Coast Based on Compressive Sensing

KEY WORDS:

Hyperspectral imaging, Compressive Sensing, Quality analysis, Brazilian Coast.

PAGES: 69
BIG AREA: Engenharias
AREA: Engenharia Elétrica
SUBÁREA: Telecomunicações
SPECIALTY: Sistemas deTelecomunicações
SUMMARY:

Hyperspectral imaging has emerged as a new generation of technology for Earth observation and space exploration since the beginning of this millennium and widely used in various disciplines and applications. Detailed spectral information is acquired by hyperspectral imagers, which generally produce results that are impossible to obtain with multispectral images or other types of satellites. Many techniques for data quality analysis have been developed over the years, and adapted for hyperspectral imaging according to the objective, such as NDVI and NDWI. The great limitation of this type of technology is the storage of data due to the weight limitation that satellites have. In this way, compression techniques started to be used, such as \textit{Compressive Sensing} (CS) created for medical imaging, generating high quality images from radial lines in $k$-space even when the ratio of Shannon-Nyquist is not obtainable. Thus, this work combined the CS technique with quality analysis to reconstruct hyperspectral images of the Brazilian coast, based on data estimation in the Fourier domain and to quantify data quality based on real data obtained by the PRISMA satellite. The results obtained show both the signal-to-noise ratio of the reconstruction, where there is a variation between 45~dB and 330~dB compared to the original image and that even in images the value of radiant numbers below the ideal, the emphasis of information is visible according to the NDWI. These results suggest that \textit{Compressive Sensing} with prefiltering allows an improvement over the prefiltering technique for hyperspectral imaging, similarly to the improvement observed in recent years in medical imaging problems.

BANKING MEMBERS:
Interno - 1323316 - HUGERLES SALES SILVA
Externo ao Programa - 2793461 - LEONARDO AGUAYO
Externo à Instituição - ROBSON DE OLIVEIRA ALBUQUERQUE
Presidente - 1771918 - UGO SILVA DIAS