Automated extraction of roads, buildings and vegetation from multi-source data

This chapter gives an overview of novel approaches and developments in the area of automated object extraction from multi-source data, i.e. mostly images, but also laser and Synthetic Aperture Radar (SAR) data, from the ground, from the air and from space, since the ISPRS congress in Istanbul in 2004. It is organized firstly according to the major object types, namely roads and traffic, buildings and finally vegetation. Additionally, we have devoted one section to two new promising modelling techniques, namely appearance based and generative modelling. For both, an intuitive introduction and relevant literature are given. We also emphasize the trend towards stochastic modelling and automated learning. For all three, it is shown how they are employed for buildings, roads and trees. Before ending the overview with conclusions, we present the state of the art of the evaluation of object extraction approaches as this is an important prerequisite for practical implementation.