The algorithm presented performs satisfactorily for edge tracking once the first frame has been initialized with an approximately correct model. For structured roads, only a crude approximation is required because the road edge pixels are easier to detect. The separate confidence levels for the left and the right edges allows for the proper weighting of the edges when computing such parameters as heading angle, offset and curvature. The model with the higher confidence should have a higher weight. A significant weakness of the algorithm is the fact that the edges are not constrained to behave in a complimentary manner. In this case, the edge pixels for the left and right would still be detected separately as above. The final models for the frame would then be derived from both edges and the constraint. This would be an advantage when one edge is not detected due to an occlusion, for example. A future task is the selection and implementation of such a constraint. However, the algorithm presented above can still be used reliably to calculate the vehicle parameters such as the heading angle. This is the next phase of the project.
Adaptive order explicit polynomials for road edge tracking
Adaptiver Polynom-Ansatz zur Straßenrand-Verfolgung
2001
5 Seiten, 3 Bilder, 8 Quellen
Conference paper
English
Adaptive order explicit polynomials for road edge tracking
British Library Conference Proceedings | 2001
|Explicit Traffic Wave Tracking
Online Contents | 2008
|Style adaptive contour tracking of human gait using explicit manifold models
British Library Online Contents | 2012
|