Visão Computacional

1.  Representação de 
     imagens
2.  Filtragem de imagens
3.  Detecção de Bordas
4.  Segmentação Simples
5.  Crescimento de Regiões
6.  Segmentação com
     Filtros
7.  Segmentação a Cores
8.  Análise de Texturas
9.  Análise de Texturas
     Multiescalar 
10. Redes Neurais
11. Morfologia Matemática
12. Convolução
13. Esqueletonização
14. Técnicas Estatísticas
15. Fractais
16. Reconhecimento de
      Formas
17. Representação de 
      Objetos
18. Quadtrees e Octrees
19. Visão Estereo
20. Inteligência Artificial
21. Controle de qualidade
22. Robótica
23. Medicina
24. Sensoriamento remoto

Prof. Aldo von Wangenheim 
Currículo...
Publicações
Pesquisa
Projetos
Ensino de Graduação
Ensino de Pós Graduação 
Cursos

Você lê?
 

1. Técnicas de  Sensoriamento Remoto
   Apresentação de Fábio Alexandrini, 2001

Abaixo você vai conhecer algumas técnicas e alguns exemplos muito interessantes. Veja aqui como são tomadas imagens de radar estereoscópicas e como se realiza a análise destas imagens para a geração de mapas topográficos e a descoberta de plantações, desertos, etc.

A área da análise de imagens do alto inclui não somente imagens de satélites, mas também imagens tomadas a partir de aviões, técnica denominada aerolevantamento.
 

Exemplos Abaixo: Imagens de radar tomadas  a bordo do Space Shuttle pelo German Remote Sensing Data Center, uma iniciativa do Centro Alemão de Pesquisas Aeroespaciais - DLR.

1. Interferometria por Radar
2. Mapas de Coerência
3. Interpretação de Imagens de Satélites
4. Como se constrói um Mapa de Elevação
5. Imagens de Interferometria de Paranaguá, PR, Brasil

It is not possible to measure the three-dimensional position of a point on the ground using only one radar image. Therefore, in analogy to optical stereoscopy, two images taken from different positions are combined. 

Isto pode ser realizado através da intefrometria, que pode ser de passada única ou de dupla passada.

The Shuttle Radar Topography Mission of DLR allows to record both images simultaneously using two antennas. The second antenna is placed at the tip of a 60 metres long mast. 

The inherent similarity of the two images, the coherence, is measured and serves as indicator of the expected quality of the derived digital elevation model. 

Mapas de Coerência

The similarity between the two radar images is determined using the so called coherence. The choherence values can also be used to indicate the expected quality of the Digital Elevation Model. In the coherence map for every pixel the similarity of the 

two images is shown as a grey scale value. Coherent areas are shown bright, areas with a low coherence like open water are represented in dark colours. It is not possible to determine altitude values of in-coherent areas. Coherence maps are beeing sucessfuly used for vegetation- classification. In order to get an interferogram both images have to be fitted exactly to each other and filtered to enhance the image quality. Every colour value represents a phase value between 0 and 2pi, where the colours repeat to form so-called fringes (stripes). As can be seen in the image below, the topography of the area is already visible in the interferogram. To emphasis this, the image the Digital Elevation Model was overlain on the interferogram. An shaded relief model, coloured according to the height values was used. The phase differences together with orbit information of the two antennas are used to calculate height values. In a last step those values are transformed to a map projection (UTM) or geographic coordinates. 

Imagem 1: Mapa de elevação digital da região de White Sands, EUA, artificialmente colorido de acordo com o tipo de estrutura gerado a partir da análise de imagens por radar do espaço. Mostra a localização dos exemplos abaixo

The Digital Elevation Model - DEM is complemented by information characterizing the shape and the physical properties of the imaged objects and land cover structures, making an evaluation and characterization of inhabited settlements possible, providing information for agriculture and micro farming, mapping the humidity and roughness of soil. The following examples illustrate results of advanced methods for image information retrieval which can help users to find relevant information in large sets of DEM and image data, to fuse the information and to understand the scenes.
 

Como se constrói um Mapa de Elevação Interpretado ? 
 

	Image 2: High resolution image of the investigation area and surroundings. The speckle, an inherent effect of SAR images, hinders the evaluation of details, both structural and radiometric.
	Image 3: After application of advanced methods for information extraction an image with high quality details is obtained.
	Image 4: Example of an area with rich geomorphometric information.
	Image 5: Landform classification of the DEM of the region presented in image 4. Geomorphometric applications are based on information extraction from DEM data. The discrimination of landforms, i.e. the classification of plains and tablelands, open hills, hills and mountains, can be performed by aggregation of information of terrain height and terrain structures as parameters of a Gasuss-Markov random field. The result of a supervised classification based on interactively learned examples gives results which explain complex structures of the Earth surface.
	Image 6: Example of scene with relevant radiometric information.
	Image 7: The classification of land structures in Image 6 is a difficult task due to the speckle process and the influence of topography. The same image "intensity" is recorded for irrigated agriculture fields or the slopes of the mountains. The fusion of information on backscatter and texture allows the separation of the relevant image regions.

Contato: Tel.: +55-48-331 7552/9498 FAX: +55-48-331-9770 awangenh@inf.ufsc.br