Application of Mathematical Morphology and Markov Random Field Theory to the Automatic Extraction of Linear Features in Airborne Images

Abstract

In this paper we present a model-based approach to the automatic extraction of linear features, like roads and paths, from aerial optical images. The proposed method consists of two steps. The first step utilizes local information related to the geometry and radiometry of the structures to be extracted. It consists of a series of morphological filtering stages. The resulting image (response) serves as input to a line-following algorithm, which produces a set of line segments. In the second step, a segment linking process is carried out incorporating contextual, a priori knowledge about the road shape, with the use of Markov random field (MRF) theory. In this approach the extracted line segments, produced by the morphological operators, are organized as a graph. The linking of these segments is then achieved through assigning labels to the nodes of the graph, using domain knowledge, extracted line segments measurements and spatial relationships between the various line segments. The interpretation labels are modeled as a MRF on the corresponding graph and the linear feature identification problem is formulated as a maximum a posteriori (MAP) estimation rule. The proposed approach has been successfully applied to airborne images of different profile