This research presents an approach of automatic extraction of the stripes that signify the borders of highway lanes from large scale digital images. The highway width and centerline can be resolved consequently. In this approach, edge points yielded by employing canny edge operator are aggregated into edge segments. Then, non-parametric figures are generated and only closed shapes are extracted out of them. At last, the border stripes are recognized by applying relevant shape measures to all identified closed figures. Two image patches, extracted from a large-scale digital aerial image are used for the research experimentation. Each of the two patches exhibit part of a highway network. The results reached by applying the proposed approach have confirmed the validity of utilizing appropriate shape measures in automatic recognition closed figures. |
