Autoria: GAMBA, S. R. H.; SANO, E. E.

Resumo: ABSTRACT -This study deals with the identification of vessels in airborne synthetic aperture radar (SAR) images. The objective is to identify the optimal GIS-based integration approaches, image enhancements, morphological filters, classifiers and processors that enable better identification of ships in SAR images from the coastal areas of Brazil. The methodology included the analysis of five digital images from three study areas (Port of Tubarão (Es), Port of Santos (SP), and Snake Island (RS)) were exported to MS Excel? spreadsheet and statistical packages SPSS? and MINITAB? to be analyzed statistically. The images were further processed using ENVI 4.5 on different highlights (2% linear, Gaussian, equalization, square root and contrast from 50 to 200), morphological filters (dilation, erosion, opening and closing), non-supervised classifiers (ISODATA and Kmeans clustering), supervised classifiers (parallelepiped, minimum distance, Mahalanobis distance, maximum likelihood, spectral angle map, divergence of spectral information, binary encoding and support vector machine) and processors (by decorrelation highlight, saturation and synthetic color image). Results of this study showed that the the most appropriate SAR image to identify vessels was the L-band with HH, VV and VH, or HH, VV and HV polarizations, followed by application of contrast enhancement of 50-200, morphological opening filter and classifier support vector machine or synthetic color image processor.

Ano de publicação: 2011

Tipo de publicação: Artigo em anais e proceedings

Unidade: Embrapa Cerrados

Palavras-chave: Porto Marítimo, Radar, Remote sensing, Sensoriamento Remoto