Automatic extraction of the road network from THRS images

In this work, we address the problem of road network extraction from THR satellite images and its implementation with OTB. More precisely, it will be a question for us to study and implement under OTB, a method of extraction of road networks existing in the literature. In this paper, we have presented a processing chain consisting of three blocks: the first one is dedicated to the multiscale analysis for the reduction of the noise effect, the second one to the extraction of roads by the spectral angle technique and to the tracking by the kalman filter and the last one subsamples the resulting image by decimating it by a factor of 2. These three blocks are executed in a loop for the pyramidal generation. Its implementation under OTB has been facilitated by the multitude of applications it offers and their easy integration for a specific application. This methodology takes into account radiometric and geometric information and corrects the shortcomings of several existing systems, notably by considerably reducing the false alarm rate. This book is intended for geomaticians, remote sensors and topographers.