Abstract:
This paper presents a comprehensive approach to effectively extract cartographic urban data from high resolution satellite imagery. It
consists of a sequence of image processing techniques, for image segmentation, based on RGB band separation, analysis and preprocessing,
followed by a morphological-based approach for data segmentation. The chosen image objects for this study are roof-tile
buildings and road network. The first step of this approach consists of a spectral response evaluation for roof-tile building objects in
a dense urban environment, being those enhanced through proper sequence of standard arithmetic operators, applied to RGB bands,
segmented and generalized. The second step aims at urban road network segmentation for cartographic representation purposes, by
combining watershed and dual reconstruction morphological transformations, which characterize the hierarchical waterfall concept.
For the latter concept, a new approach is developed in order to improve hierarchical segmentation procedure, to better induce object
discrimination. Each one of the referred objects will be segmented in separate. The road network segmentation will have in
consideration the previous result of roof-tile buildings extraction. Finally, segmented objects will be compared with other extraction
results in order to do proper validation. The method is applied over high-resolution ortho-rectified images, taken from satellite, of the
city of Lisbon. Results will show the practicality of the method for purposes of cartographic data structures acquisition, from highresolution
satellite imagery, aiming urban management and GIS applications.