Title :
Recursive surface structure recognition for safe navigation
Author_Institution :
Univ. der Bundeswehr Munchen, Neubiberg, Germany
Abstract :
Methods for extracting in real time the surface structure in a certain look-ahead range from image sequences while driving are reviewed. Low spatial frequency elevation profiles of roads are the only ones successfully recovered in real-time up to now. Low frequency means that the characteristic geometrical dimension for curvature estimation is at least an order of magnitude larger than the axle distance of the vehicle carrying the camera. For the more general case of a surface with relatively high spatial frequencies an approach for real-time recognition is proposed which separates low, medium and high spatial frequency components; these components are estimated in a sequential, iteratively parallel manner. The formulation of the approach is given
Keywords :
computational geometry; computer vision; iterative methods; navigation; object recognition; parallel processing; recursive estimation; road vehicles; stereo image processing; 3D structure recognition; computer vision; geometric model; image sequences; iterative method; parallel processing; real time systems; recursive surface structure recognition; safe navigation; spatial frequency; stereo camera system; Axles; Cameras; Frequency estimation; Geometry; Navigation; Roads; Shape; Solid modeling; Surface structures; Vehicles;
Conference_Titel :
Intelligent Vehicles '95 Symposium., Proceedings of the
Conference_Location :
Detroit, MI
Print_ISBN :
0-7803-2983-X
DOI :
10.1109/IVS.1995.528319