Segmentação morfológica aplicada à cartografia
Carregando...
Data
2003
Autores
Medeiros, Nilcilene das Graças [UNESP]
Título da Revista
ISSN da Revista
Título de Volume
Editor
Universidade Estadual Paulista (Unesp)
Resumo
A segmentação de imagens é um processo muito utilizado e relevante em diversas áreas do conhecimento. Neste trabalho tem-se um exemplo na área de cartografia. O processo de extração de feições não é uma tarefa fácil e necessita do uso de técnicas mais complexas e de dados com boa qualidade para o sucesso dos resultados. Assim, este trabalho teve por objetivo reunir técnicas de Processamento Digital de Imagens (PDI), em especial a Morfologia Matemática, em conjunto com imagens de Sensoriamento Remoto, na utilização específica da segmentação de imagens, visando a obtenção de resultados relevantes à área de cartografia. Visto que os operadores morfológicos utilizam como padrão de reconhecimento apenas os valores de brilho, a idéia foi realizar um tratamento de realce que melhorasse a qualidade das feições de interesse para serem extraídas no processo de segmentação. A melhora da qualidade dessas imagens pode contribuir para a obtenção de resultados mais significativos. Para isto, foram utilizados alguns dos operadores morfológicos contidos na toolbox de morfologia matemática acoplada ao software MATLAB, voltados a etapa de préprocessamento de imagens originais (“brutas”) e a implementação de duas abordagens utilizadas para a geração da imagem de marcadores. A imagem original e/ou tratada e a imagem de marcadores são usadas como parâmetros de entrada no processo de segmentação com o operador watershed. Os operadores implementados foram o gradiente morfológico multi-escala e o mínimos impostos. As imagens utilizadas são imagens orbitais do satélite Landsat, dos sensores TM e ETM+, disponíveis no banco de imagens da FCT - Unesp. Após a realização dos testes foi verificado o desempenho dos operadores implementados e realizada a análise dos resultados...
Image segmentation is a process often used in several different areas including Cartography. Feature extraction is a very troublesome task, and successful results require more complex techniques and good quality data. The aim of this work is to study Digital Image Processing techniques, with emphasis in Mathematical Morphology, using Remote Sensing imagery. These techniques were used to segment images aiming significant cartographic applications. Considering that morphological operators use grays levels as the major recognition pattern, the proposal of this research is to apply previously an enhancement technique aiming the improvement of the interest features to be extracted in the segmentation process. More meaningful results can be attained with this previous image enhancement step. In order to test the proposed approach some morphological operators related to early processing were applied in unprocessed images using an existing MatLab Toolbox. Furthermore, two distinct techniques generating marker images were implemented in the MatLab environment. The raw and the enhanced images, and the marker image were used as input parameters in the segmentation process with the Watershed operator. The implemented operators were the morphological multi-scale gradient and minima imposed. Orbital images of the Landsat satellite, sensors TM and ETM+ , available in the image database of FCT-UNESP were used. The performance of the implemented operators and the results were analised after some tests with the existing images. Extraction of linear features, such as highways, airports, etc., can contribute in cartographic applications as the feature identification or cartographic updating, considering that orbital images contain more recent information and the changes can be emphasized and extracted. The proposed objectives were reached with contributions to the segmentation processes using Watersheds.
Image segmentation is a process often used in several different areas including Cartography. Feature extraction is a very troublesome task, and successful results require more complex techniques and good quality data. The aim of this work is to study Digital Image Processing techniques, with emphasis in Mathematical Morphology, using Remote Sensing imagery. These techniques were used to segment images aiming significant cartographic applications. Considering that morphological operators use grays levels as the major recognition pattern, the proposal of this research is to apply previously an enhancement technique aiming the improvement of the interest features to be extracted in the segmentation process. More meaningful results can be attained with this previous image enhancement step. In order to test the proposed approach some morphological operators related to early processing were applied in unprocessed images using an existing MatLab Toolbox. Furthermore, two distinct techniques generating marker images were implemented in the MatLab environment. The raw and the enhanced images, and the marker image were used as input parameters in the segmentation process with the Watershed operator. The implemented operators were the morphological multi-scale gradient and minima imposed. Orbital images of the Landsat satellite, sensors TM and ETM+ , available in the image database of FCT-UNESP were used. The performance of the implemented operators and the results were analised after some tests with the existing images. Extraction of linear features, such as highways, airports, etc., can contribute in cartographic applications as the feature identification or cartographic updating, considering that orbital images contain more recent information and the changes can be emphasized and extracted. The proposed objectives were reached with contributions to the segmentation processes using Watersheds.
Descrição
Palavras-chave
Cartografia, Processamento de imagens - Tecnicas digitais, Computação de imagens, Segmentação de imagens, Imagens orbitais, Atualização cartográfica, Images orbital, Computer Images
Como citar
MEDEIROS, Nilcilene das Graças. Segmentação morfológica aplicada à cartografia. 2003. 131 f. Dissertação (mestrado) - Universidade Estadual Paulista, Faculdade de Ciências e Tecnologia, 2003.