Spatial-Temporal Junction Extraction and Semantic Interpretation

Abstract. This article describes a novel junction descriptor that encodes junctions’ semantic information in terms incoming lines’ orientations, both in 2D and 3D. A Kalman filter process is used to reduce the effect of local noise on the descriptor’s error and to track the features. The improvement gained by our algorithm is demonstrated quantitatively on synthetic scenes and qualitatively on real scenes.

• Type: Conference
• Year: 2009
• Where: ISVC
• Authors: Kasper Broegaard Simonsen, Mads Thorsted Nielsen, Florian Pilz, Norbert Krüger, Nicolas Pugeault
• Keywords: Applied Computing, Descriptor’s Error, ISVC 2009, Kalman filter Process, Terms Incoming Lines