Detection and classification of painted road objects for intersection assistance applications

Abstract

For a Driving Assistance System dedicated to intersection safety, knowledge about the structure and position of the intersection is essential, and detecting the painted road signs can greatly improve this knowledge. This paper describes a method for detection, measurement and classification of painted road objects that are typically found in European intersections. The features of the painted objects are first extracted using dark light dark transition detection on horizontal line regions, and then are refined using gray level segmentation based on Gaussian mixtures. The 3D bounding box of the objects is reconstructed using perspective geometry. The objects are classified based on a restricted set of features, using a decision tree and size constraints.