Semiautomatic extraction strategy of urban road junctions from high resolution orthoimages

Abstract

The aim of this paper is present and asses a strategy for semi-automatic extraction of road junctions from high resolution orthoimages take over urban areas. The proposed strategy is performed by collecting a set of samples data training and generates decision tree containing the explicit knowledge of the object classes, by WEKA software. On the roads defined, are applied morphological operators and algorithms that allow the detection of skeletons, junctions, selection of reference points and the roads junction extraction. The proposed strategy was tested with real data and the results were analyzed, allowing an evaluation of the strategy, as well as potentially problematic situations. The results presented in the experiments demonstrated the feasibility of road extraction in urban area scenes, road networks formed by simple and complex intersections.