A Morphological Methodology for Features Identification in Satellite Images for Semi-automatic Cartographic Updating

摘要

This work investigates a way to exploit the information of satellite images in order to identify cartographic features, aiming at developing a software tool able to update digital maps automatically. A cartographic feature, like any object present in a multi-channel image, is a set of pixels with similar spectral response and a certain spatial relation between them. The current algorithm works iteratively and mixes the spatial information with the spectral one in an appropriate way to finally detect the whole shape of a cartographic feature starting from a pixel marked previously by the user in a remotely sensed image. It is also shown that Mathematical Morphology (MM) operators can handle the spatial and spectral information decreasing the computational cost. First, the structure of the main algorithm is presented, showing each step of its operational sequence. Then, some application examples are reported and, finally, some remarks illustrate the future possibilities of implementation and development of the algorithm.