An object-oriented environment for linear feature extraction from aerial images.

Abstract

We propose a new method for the detection of road networks in SPOT$\sp1$ images, with possible applications in automated cartography. This method is based on the integration of image segmentation techniques, in particular the watershed and Duda transformations, in a parallel Single Instruction Stream Multiple Data Stream (SIMD) environment. The Duda Road Operator provides the marking function for the watershed transformation of the image. Examples involving SPOT images are used and the resulting road networks are shown. ftn$\sp1$Systeme Probatoire d'Observation de la Terre.