Lane Boundary Detection Using Statistical Criteria

This paper describes an approach of lane boundary detection especially for country roads elds by artiicial vision. It uses statistical criteria such as energy, homogeneity, contrast etc. to distinguish between road and non-road-area. To model the borders of the road and to accelerate the method, we combine it with a model based image segmentation technique. The model is adapted to the found lane boundaries by a direct estimation of the model parameters which is based on chi-square tting. Additionally we will show that the grey value statistics for video images of ourdoor scenes are independent from the estimation direction for small distances.